Patent Grant
10182258
Advertising forms an important part of broadcast programming including broadcast video (television), radio and printed media. The revenues generated from advertisers subsidize and in some cases pay entirely for programming received by subscribers. For example, over the air broadcast programming (non-cable television) is provided entirely free to the subscribers and is essentially paid for by the advertisements placed in the shows that are watched. Even in cable television systems and satellite-based systems, the revenues from advertisements subsidize the cost of the programming, and were it not for advertisements, the monthly subscription rates for cable television would be many times higher than at present. Radio similarly offers free programming based on payments for advertising. The low cost of newspapers and magazines is based on the subsidization of the cost of reporting, printing and distribution from the advertising revenues.
Along with the multitude of programming choices that the television viewer faces, the viewers are subject to advertisements. While advertisements are sometimes beneficial to subscribers and deliver desired information regarding specific products or services, consumers generally view advertising as a “necessary evil” for broadcast-type entertainment. A prior art (present model) of providing advertisements along with actual programming is based on linked sponsorship. In the linked sponsorship model, the advertisements are inserted into the actual programming based on the contents of the programming, e.g., a baby stroller advertisement may be inserted into a parenting program. Even with linked sponsorship, advertising, and in particular broadcast television advertising, is mostly ineffective. That is, a large percentage, if not the majority, of advertisements do not have a high probability of affecting a sale. In addition to this fact, many advertisements are not even seen/heard by the subscriber who may mute the sound, change channels, or simply leave the room during a commercial break. The reasons for such ineffectiveness are due to the fact that the displayed advertisements are not targeted to the subscribers' needs, likes or preferences. Generally, the same advertisements are displayed to all the subscribers irrespective of the needs and preferences of the subscribers.
In order to deliver more targeted programming and advertising to subscribers, it is necessary to understand their likes and dislikes to a greater extent than is presently done today. Targeting of an ad requires knowing certain attributes of the target viewer, demographic, psychograph, and any data relevant to determining the relative appropriateness of an ad for the particular viewer. Systems which identify subscriber preferences based on their purchases and responses to questionnaires allow for the targeted marketing of literature in the mail, but do not in any sense allow for the rapid and precise delivery of programming and advertising which is known to have a high probability of acceptance to the subscriber. In order to determine which programming or advertising is appropriate for the subscriber, knowledge of that subscriber and of the subscriber's programming preferences is required. Characterizing or profiling viewers based on viewing habits may be used to achieve targeted advertising.
Methods for monitoring the viewing habits of television viewers, for classifying TV programming into categories, and for using the viewing habits for determining viewing preferences have been previously disclosed. For example
The following detailed description will be better understood when read in conjunction with the appended drawings, in which there is shown one or more of the multiple embodiments of the present invention. It should be understood, however, that the various embodiments of the present invention are not limited to the precise arrangements and instrumentalities shown in the drawings.
A method for delivering targeted ads to different demographic groups in a television environment was disclosed by Wachob (U.S. Pat. No. 5,155,591), assigned to General Instrument of Hatboro, Pa. Wachob discloses a cable television system for broadcasting different commercial messages to different demographically targeted audiences. Demographic information is obtained and targeted audiences are formed based on subscriber address (i.e., geographic location) or on household survey information such as a viewing habit diary kept by the subscriber.
Methods for delivering advertising or other customized programming to viewers in a television environment based on previous viewing habits or menu selections has previously been disclosed. For example, Despain et al. (WIPO publication WO 00/14951), assigned to Next Century Media, Inc. of New Paltz, N.Y., discloses a system for targeted advertising in a digital system. In the system, a digital set-top box (STB) captures and uploads household data viewing preferences to the cable operator head end, which can then be used to deliver targeted ads and other content to the viewer based on the viewer preferences. An on-screen questionnaire is used to elicit demographic attributes and preferences from viewers. “Boo” and “applause” buttons on the remote control are used by the viewer to indicate viewer likes/dislikes. Channel change data can be captured and sent upstream, and in conjunction with data from the questionnaires and interactive buttons, may be used to provide each viewer with their own custom menu and forms of customized programming. However, Despain et al. do not teach local storing and processing of data to generate a viewer profile to be stored and utilized at the STB. Moreover, it does not teach specific methods on how to create a demographic, psychographic or other viewer profile from a multitude of viewer interaction data or how to correlate or use those profiles for the delivery of targeted advertisements.
A preference agent for monitoring television programs watched by a viewer is disclosed in Gogoi et al. (WIPO publication WO 99/65237), assigned to Metabyte, Inc. of Freemont, Calif. The preference agent, located within a STB or PVR, also retrieves category information about viewed programs, and generates a viewer program preference profile. The preference agent automatically records or suggests programs of interest to the viewer based on the viewer's program preference profile.
The creation of user profiles on an interactive computer is disclosed in Freeman et al. (U.S. Pat. No. 5,861,881), assigned to ACTV Inc., of Freeman, N.Y. The profiles are based on the selections made during one of the interactive programs, however, are limited to interactive program activity by the viewer, and are not based on general viewing habits or general surf activity.
Hendricks et al. (U.S. Pat. No. 6,160,989), assigned to Discovery Communications Inc., of Bethesda, Md., discloses a network controller that provides monitoring and control of STBs. The network controller also gathers data received from the STBs to compile subscriber viewing information and programs watched information. The data is processed to generate packages of advertisements, as well as account and billing reports, targeted towards each STB. To build a personal profile, the viewer answers a series of questions presented on a series of questionnaire menu screens.
Barton et al. (WIPO publication WO 00/59223), assigned to TIVO Inc. of Alviso, Calif., disclose a data storage and scheduling system, wherein viewing preferences can be inferred from viewing patterns, and where the navigation actions of the TV channels by the viewer are recorded, stored, and then sent upstream. However, Barton et al. do not teach or suggest local profiling or viewer identification, or how profiles could be utilized to delivery targeted advertising.
For the foregoing reasons, a need exists for a method and system for monitoring click-stream and other interactivity of a viewer with the viewer's television viewing environment and generating one more viewer profiles therefrom. Additionally, a need exists for the monitoring of interactivity and generation of viewer profiles to be performed within the television viewing environment (i.e., TV, STB, PVR). Furthermore, a need exists for such profiling to be done in a secure and privacy-protected manner. Moreover, a need exists for a reliable way of automatically, detecting or inferring, which specific individual or individuals, are actually watching the TV in a household comprising more than one individual at a particular time, and for generating one or more profiles per each individual.
The invention comprises a method of characterizing or profiling one or more viewer's, by monitoring and processing at the viewer set-top or receiver, each viewer's interactivity (e.g., via a remote control unit) with the set-top receiver, and then generating one or more profiles for each viewer based on one or more of the multitude of interactions of each viewer and on, in general, the viewing habits and preferences of the viewer. The invention further comprises a method of automatically and reliably, detecting or inferring at a particular time, which specific individual or individuals, are actually watching (i.e., interacting with the remote) the TV in household comprising more than one individual. Such viewer identification and profile generation can be used to facilitate the delivery of targeted content including targeted advertising.
The invention also comprises a system for carrying out the local (i.e., at the set-top) profiling of viewers based on their viewing preferences and other interactions at the set-top and for distinguishing between viewers in a multi-viewer household. The invention further comprises a profiler or profiling agent, resident at the viewer's receiver or set-top, and whereby the click-stream and other interactivity by a viewer can be monitored and processed to generate one more viewer profiles and for distinguishing between viewers in a multi-viewer household.
In a preferred embodiment, the profiling agent is resident at a viewer's receiver and is responsible for profiling a single household, and distinct individual members of that household. The set-top box profiler monitors the viewing behavior and other receiver interactivity of the viewers including their viewing preferences and utilizing data about programming viewed or not viewed, derives characteristics about and generates profiles of the household and of individual viewers within the household. In a preferred embodiment, the profiles are comprised of profile categories, including the categories of preferred programs, preferred networks, viewing duration, channel change frequency, and holding factor per program or program category. The program data may be delivered to the set-top box periodically (e.g., on a carousel) or may be delivered with the video (e.g., in the VBI or in the PSI).
In a preferred embodiment, the profiler identifies different viewing sessions, a period of time during which viewership is static and does not change (i.e., a session of one or more specific viewers), based on viewing and other habits and preferences exhibited by the viewer or viewers, and creates session profiles for each session. Signature profiles are created from one or more correlated session profiles, and subsequent session profiles are correlated and matched with existing signature profiles. Sessions (or session durations) may be defined by, for example, the cycling of power (on/off) of the set-top box, a specific window of time or day-part, periods of inactivity, and monitoring channel change, viewing and other viewer activity.
In one embodiment, the profiler comprises an event queue that stores viewer interactivity as viewer-generated events. The profile engine accepts events from the event queue, reads database information (e.g., program data), and processes the events to produce the subscriber profiles. Events are dispatched to the profiling engine based on the clock time, and each of these events may be used to update and modify the viewer's profile. In a preferred embodiment the profile engine uses filters to process events. Each filter may handle one or more profiles elements, determining whether or not each specific event is applicable to (i.e., should be used to update) one or more profiles.
In one embodiment, automatic session detection uses channel change and other information to dynamically determine which viewer is watching television at any given time. In a preferred embodiment, reliable viewer identification is accomplished and more complete, updated and accurate profiles of particular viewers are generated, by combining current session data with historical data for those viewers. This historical data, or signature, is an aggregation of session data for a particular viewer or type of viewer, and reflects a set of viewing and interactivity habits. During a specific session, as the profiler collects viewer interactivity session data, the profiler continuously correlates the session data with existing signature data in order to match the session data to a specific signature. This correlation associates long term viewing habits with a particular user based on their short term viewing habits. Moreover, matching session data with a signature identifies a particular viewer and an associated profile.
In one embodiment, the signature profile represents an individual viewer. In an alternate embodiment, the signature profile represents a plurality of household viewers. In another embodiment, the signature profile is a standard profile, created independently of any session data or created from aggregated session data from a plurality of household.
Default or standard signature profiles, comprising standard profile categories may be generated, independent of local viewing sessions, and may be downloaded to the local receiver or set top. Local viewing session profiles are correlated with these standard signature profiles to identify one or more particular attributes of the session viewer.
Advantages of the invention include the ability to, based on viewing habits and other interactions, identify individual viewers or groups of viewers that possess or share certain preferences, demographic, or other relevant traits, and the ability to identify one or more particular viewers, from among a larger group of viewers, and to reliably identify whether that viewer is a man, a woman, or a child. The identification of specific viewers, based on each viewers current and previous interactivity with a receiver, and the characterization (profiling) of each of those viewers (or groups of viewers) creates enormous opportunities for content providers and advertisers to address their content to an individual or groups of individuals with particular demographic traits or who have particular interests. An additional advantage of the invention is that the local monitoring and profile generation at the set-top is consistent with and affords solutions to security and privacy considerations.
These and other features and objects of the invention will be more fully understood from the following detailed description of the preferred embodiments that should be read in light of the accompanying drawings.
The accompanying drawings, which are incorporated in and form a part of the specification, illustrate the embodiments of the present invention and, together with the description serve to explain the principles of the invention.
In the Drawings:
In describing a preferred embodiment of the invention illustrated in the drawings, specific terminology will be used for the sake of clarity. However, the invention is not intended to be limited to the specific terms so selected, and it is to be understood that each specific term includes all technical equivalents which operate in a similar manner to accomplish a similar purpose.
With reference to the drawings, in general, and
The present invention can be implemented for use with various television (TV) delivery systems including, but not limited to, digital broadcast satellite (DBS) systems, switched digital video (SDV) systems, local multipoint distribution systems (LMDS), multichannel multipoint distribution systems (MMDS), hybrid fiber coax (HFC) systems, the Internet, other cable TV (CTV) systems, or other terrestrial wireless networks. The TV delivery system can deliver programming in various forms, including but not limited to digital video, analog video, or streaming media. The programming may be compressed in accordance with a variety of now known or later discovered compression standards, such as the current Motion Picture Expert Group (MPEG-2) standard for digital video.
A CTV network 120, such as digital cable network, transmits multiple channels of TV information from a head end or central office (HE/CO) 190 via a cable network 122. Particularly, the channels are transmitted via cables 124, such as fiber optic cables, to nodes 126. The nodes are essentially switching/routing stations that service multiple homes (usually a few hundred). The nodes 126 route the signals to individual subscribers 128. For digital cable, the individual subscribers 128 will have STBs 130 that select a particular channel from the transmit stream, demodulate it and forward it for display on one or more monitors (i.e., televisions) 132. Upstream information may be sent from the STB 130 to the HE/CO 190 via a dedicated upstream channel over the cable. In cable systems that do not support two-way communication, the upstream “channel” can be through the telephone as described above in connection with DBS systems 100.
A VDSL system 150 transmits TV programming over the regular telephone network. Particularly, TV signals are transmitted from a broadband distribution terminal (BDT) 152 within the HE/CO 190 via cables 154, such as fiber optic cables, to a universal service access multiplexer (USAM) 156 that delivers the data to multiple individual subscriber households 160 via regular telephone twisted wire pair 158 using VDSL modems and protocols. The USAM 156 receives a wide bandwidth signal comprising some or all of the television channels. However, because of the bandwidth limitations of twisted pair wire, typically only about one channel of television programming at a time can be delivered from the USAM 156 to the household. Accordingly, the subscriber has a STB 162 that is similar in functionality to the previously discussed STBs for DBS and CTV, except that when the user changes channels such as by operating a remote control, the remote channel change signal is received by the STB 162 and transmitted to the USAM 156 which switches the channel for the user and begins sending the newly selected channel to the household. Such systems are known as SDV systems. SDV systems are essentially fully modern asynchronous two-way communication networks. Accordingly, the STB 162 can transmit information upstream via the same VDSL modem that receives the downstream signals. SDV systems typically operate using an asynchronous transfer mode (ATM) protocol that is well-known in the networking arts.
In an alternative embodiment, the TV signals are transmitted from the BDT 152 to a broadband network unit (BNU) 164. The BNU 164 delivers the data to individual households 160 using coaxial cable 166.
It should be noted that STBs are described above with reference to
The STB 220 receives program content from the HE 210 and delivers the program content to the TV, receives and processes commands from a viewer, and monitors the interactions of the viewer to generate viewer profile(s). The STB 220 includes a communications manager 222, a user interface 224, a profile engine 230, profile filters 240, a profile database 250, a program database 260, a clock 270, and an event queue 280.
The communications manager 222 handles the communications with the HE 210. The communications manager 222 may receive program content or database downloads (i.e., program data, channel maps) from the HE 210 and may transmit commands (i.e., channel changes), profiles (i.e., updated), or other information including anonymous system statistics (i.e., audience measurements). The user interface 224 allows the viewer to interact with the STB 220, for example, via a conventional remote control unit. The viewer interactions include, but are not limited to, channel and volume changes, EPG activity, and participation in interactive entertainment and advertisements.
The program database 260 stores program data including, but not limited to, name, network, time, and genre of programs being transmitted or to be transmitted from the HE 210. The event queue 280 stores both viewer-generated events and internal events. Viewer-generated events include viewer interactions as mentioned above. Internal events include, but are not limited to, the end of a program or the end of a day part. The clock 270 runs independently within the system. The clock 270 is used to mark the time that the viewer generated events occur and to trigger internal events, such as, program changes, end of day parts, and change of day. The events are dispatched from the event queue 280 to the profiling engine 230 based on the clock 270.
The profile engine 230 accepts events from the event queue 280 and, if applicable, retrieves program data related to a selected program from the program database 260. The profile engine 230 in conjunction with the profile filters 240, processes the events to produce the viewer profiles 199 which are then stored in the profile database 250. Each of the profile filters 240 handles a single profile element (as will be described further herein).
It should be noted that typical STBs likely include some type of user interface and network interface (i.e., communications manager). The communications manager 222 and the user interface 224 may be typical components of a STB or slight modifications thereof. Also, the clock 270 may be standard component of a STB. The profile database 250, the program database 260 and the event queue 280 may be a single memory device or may be separate memory devices. The profile engine 230 and the profile filters are resident software applications stored and running on the STB 220. As one of ordinary skill in the art would know, there are numerous variations of this architecture that are well within the scope of the current invention.
The term program data in the context of the present invention is meant to include and encompass one or more subsets of information, which identifies, describes and generally characterizes specific television programs and television networks, categories of programs and networks, etc. Program data can be readily obtained from several commercial enterprises including TVData of Glen Falls, N.Y. In a preferred embodiment, program data is used in order to be able to characterize what program and networks are viewed, surfed, or the like. For instance, program data is used to convert channel, date, and time into a program name and content type. This information determines the types of programs that are preferred by a household and how long certain programs hold their attention.
In a preferred embodiment, the HE 210, or other upstream system, generates and stores the program database 214 and transmits relevant program data to the STB 220. According to one embodiment, the relevant program data are placed on a download carousel, where they are transmitted in their entirety to the STB 220. Alternatively, the relevant program data can be downloaded periodically in smaller increments. In another embodiment, the program data is transmitted with the programs. For example, the program data is transmitted within the vertical blanking interval (VBI) or the program specific information (PSI) of a transport stream, such as, a Motion Picture Expert Group (MPEG-2) transport stream. It should be understood however, that many alternate methods for getting the program data exist and the utility and scope of the present invention is not limited by how the data is obtained.
In one embodiment, the HE 210 aggregates the program data by cable supplier, date, time, and television network to produce a variety of tables. These tables or a variation thereof are downloaded to the STB 220. Furthermore, the STB 220 may receive program data in one format and subsequently process the downloaded program data to produce other tables and alternate formats. As will be evident to those skilled in the art, the format of the program data at the HE 210, the format by which it is downloaded to the STB 220, and the format of the program data stored at the STB 220 can be widely variable. The invention is in no way limited to the specific formats discussed and illustrated below.
The session profiles 294 identify characteristics about the user (or group of users) for that particular viewing session. The signature profiles 295 are a compilation of closely related session profiles 294 (are associated with the same user or group of users). The VCPS 290 can identify which user (or group of users) are interacting with the TV by correlating the current session profile 294 with one or more signature profiles 295. The session profile 294 is identified with the signature profile 295 having the highest correlation, as long as it meets a predefined correlation threshold. The identification is an identification of preferences and not necessarily an actual identification of the user (or users). The viewer profiles 293 can be used for a variety of purposes including, but not limited to, delivering targeted content including advertising and for distinguishing a particular viewer from a household of multiple viewers.
As would be understood by those of ordinary skill in the art, the STB 220 will likely not have sufficient memory to store the entire program database 214 in the program database 260. As such, either the HE 210 will modify the program database 214 and transmit the modified database to the STB 220 or the HE 210 will transmit the entire program database 214 and the STB 220 will process the data and store a modified version thereof. In either event, exemplary headers and table formats for the STB 220 are discussed below.
Within the STB 220 the channel map and the network tables can be combined into a single table that would be specific to a local service domain or zone.
A program table within the STB 220 may be either a fixed or variable length table. For a fixed length program table the time period covered will be divided into time slots of equal length, for example, 5 minutes, 10 minutes, 30 minutes. As would be obvious to one skilled in the art the granularity depends on the amount of storage in the STB 220. The fixed length program table stores information about the programs appearing in those slots for each of the networks. It is possible that one network may have multiple programs during a single time slot. In those cases, the program table captures the program that airs for the majority of the time slot. For example,
The fixed record size and fixed slot size allow the VCPS 290 to very rapidly locate the program information for the network at the given time. The fixed length program table is also easily divided into one or two hour increments. This enables the program data for the current time to be transmitted to the STBs 220 quickly and often. In a fixed length table, each network has only one entry per time slot. The program data is found by locating the start of the time slot corresponding to the event time and adding the netidx 738.
As one skilled in the art would recognize, the fixed length program table size would be approximately 96K for 200 channels having 30 minute time slots and 24 hours of data (200 channels*10 bytes/time slot/channel*48 time slots). If the program table size was too large, the time slots could be increased, the amount of time captured for in the table could be decreased, the netref 930 could be removed so that the verification wasn't performed or other modifications that would be obvious to one of ordinary skill in the art.
A variable length program table is particularly useful when the accuracy of exactly what program is being viewed is of critical importance. The variable length program table includes a variable length program header (
In a variable length table, all program data for a single network is specified prior to moving on to the next network. The netidx 738 provides the location for the start of the program data for that network. The program information is then sequentially searched to locate the program time that corresponds to the event. A variable length program table provides the VCPS 290 the most exact data about the programs. Since the number of records is variable, the exact table size is unknown. As would be obvious to one of ordinary skill in the art, the data sizes of the program records can be reduced at the cost of flexibility and accuracy.
The ClickStreamProcessor 1230 retrieves events from the ClickEventListener 1220 and, if applicable, requests program data for a selected program from the ProgramContentManager 1240. The ProgramContentManager 1240 retrieves the applicable program data from ProgramContent 1270. The program data includes, but is not limited to, program title, program genre, program category, and network. The ClickStreamProcessor 1230 then processes the events and, if applicable, program data utilizing ClickStreamFilters 1260 to filter events such that only the appropriate event types are used in generating profile categories.
The ClickStreamFilters 1260 direct and add data to the relevant profile category to generate or update a Profile 1290 comprising multiple profile categories. Such categories, as described further herein, include preferred networks, preferred programs, viewing duration, channel change frequency, and holding factor. The ClickStreamProcessor 1230 also generates a ClickEventRecord 1295, which is a record of events for a given session or other time period.
In one embodiment, the ClickStreamProcessor 1230 will continually poll the ClickEventListener 1220 for the next event. If no events have been triggered, the ClickStreamProcessor 1230 will sleep for a short period of time. Otherwise, the ClickStreamProcessor 1230 will handle an event and then immediately check for the next event. Event processing is based on the event time. According to this embodiment, the ClickStreamProcessor 1230 handles events, such as channel change and power on/off events, from the remote control to dynamically update a viewer's profile. Thus, the ClickStreamProcessor 1230 spends most of its time polling the ClickEventListener 1220 for new events. When an event occurs, the ClickStreamProcessor 1230 passes the ClickStreamFilters 1260 (an array of profile filters), which extract specific data for the household or viewer profile.
Referring back to
In a preferred embodiment, profiles generated by the VCPS 290 include multiple profile categories, each category reflecting distinct relevant characteristics about a viewer's viewing preferences and habits. Examples of profile categories include, but are not limited to:
As one of ordinary skill in the art would recognize, there are numerous other profile categories that could be included that would fall within the scope of the current invention. The current invention is in no way not limited to the categories described herein.
In order to generate a preferred programs profile, the VCPS 290 collects information (characterizations) about the programs that a viewer or household watches. The characterizations about the programs may include, but is not limited to, types, categories, genres, or some combination thereof. In a preferred embodiment, the characterizations will include genre and category for each program. The genre is a consistent high-level classification of a program (i.e., a generic set of program types or categories), such as sports, comedy, drama, etc. A category is a sub-class of the genre classification that is a more specific classification than the genre.
The program genre and categories may be defined by the VCPS 290, received from a service provider (e.g., TVData), or derived from a service provider. In a preferred embodiment, program characterizations are obtained from, for example, TVData and these characterizations are converted into genre and category. TVData provides a program type and category for each category and these characterizations are translated (e.g., mapped) into the program classifications of genre and category.
The VCPS 290 tracks the total number of seconds that a viewer watches particular program categories (genres). The VCPS 290 responds to all channel change and power on/off events. When a channel change occurs, the VCPS 290 records the time and network, and locates the program's characteristics from the program table. When the next channel change occurs, the VCPS 290 notes the elapsed time and stores that elapsed time in an array. Because it is possible that a program will end prior to a channel change, the VCPS 290 saves the elapsed time to the appropriate program category (genre) and then gets the program information for the program that is about to air such that when a channel change or power off event occurs the appropriate time spent on each program category will be accurate.
The VCPS 290 may also track the preferred program categories (genres) of the household by day or day part. A day part is a range of time during the day that is used by advertisers to characterize viewers. The VCPS 290 can create an event for a day part by saving elapsed time to the specific day part and generating a new event for the next day part.
Note that this example profile is an aggregated profile comprising more than a week's worth of data. A similar session profile comprising a much smaller span of viewing time may also be obtained by the VCPS 290. As one of ordinary skill in the art would recognize, there are numerous other formats of this graph that would be well within the scope of the current invention. For example, the bar graphs 1640 could be a pie chart or the chart could be further broken out by day or day part.
As previously discussed, program categories 1320 are more specific than genres 1310 and therefore provide for increased granularity. The VCPS 290 also tracks program categories 1320 in the same fashion as it tracks program genre 1310. An exemplary graphical representation of the program category profile is not illustrated. However, as one of ordinary skill in the art would recognize, the graph of the program category profile would be similar to the genre profile illustrated in
A viewer type profile estimates what type of viewer (i.e., man, woman or child) is watching a particular program. In one embodiment, the VCPS 290 uses program classifications (genre/category) and day part information to derive the estimates.
The VCPS 290 may adjust the probability data based on the day part. For example, because the probability that a daytime viewer is a man is lower than the probability that the viewer is a woman or a child, this fact results in an adjusted and reduced probability that the daytime viewer is male.
As one of ordinary skill in the art would recognize, applying the adjustment factor will likely mean that the sum of the probabilities for a particular day part will not equal 1.0.
The distinction of a viewer type by gender, as described above, can be readily extended to other viewer types and demographics, and in general any set of heuristic rules or probabilities can be applied to the viewer interactivity data to generate predicted viewer traits and demographics, as will be evident to those skilled in the art. The present invention is not meant to be limited to distinguishing three viewer types as described, but could include additional sets of rules and probabilities which, in conjunction with viewer interactivity data, can be used to derive or infer viewer demographics and other attributes.
In order to generate a preferred network profile, the VCPS 290 tracks the networks that are most watched by each viewer or household.
A viewing duration profile is the average duration of each viewing session. This is useful for determining how much television a viewer or household watches at a time (e.g., per session). This information can help identify the households that do not watch television often. To collect statistics for this profile, the VCPS 290 responds to power on/off events. The VCPS 290 tracks the elapsed time between a power on event and a power off event. The VCPS 290 also generates an average duration of each session (i.e., average viewing duration for all viewing sessions).
The VCPS 290 also tracks viewing duration by day or day part.
A channel change frequency profile measures how often or rapidly a viewer or household changes channels. This information can be used to determine characteristics about the household and it can help differentiate individual users within the household.
A holding factor profile generally indicates relative interest levels in certain programs by a viewer or household. A holding factor is how much of an entire program a viewer or household watches. The VCPS 290 tracks the total time that each program is viewed.
A channel search sequence is the order that the viewer typically visits specific networks, and this may be tracked by day part. The VCPS 290 tracks the order the networks are typically visited by averaging the visit position of each network over a series of channel surf sequences. This is quite useful for profiling and viewer identification as there may be certain channels that a particular viewer typically watches, or a specific surf sequence a particular viewer may perform.
Alternatively, a viewer may always sequentially search channels, such a channel range of the sequential search can help identify the viewer or an associated signature. As discussed further below, the VCPS 290 compares the session-based search order with the search order for all signatures.
The VCPS 290 keeps track of two arrays, one for the number of times the viewer visits a network (count 2560) while surfing and the other to determine the total dwell time 2570 on that network while surfing. This data can also be tracked by day part. The VCPS 290 also calculates the average channel dwell time, which is calculated by dividing the data in the total time array by the data in the total count array. The result indicates the amount of time that a viewer watched each particular network during a channel surf sequence. This result can be used to determine whether the program on that network is of interest or potentially of interest (i.e., the viewer spent time determining if they were interested) or not (i.e., the viewer was simply surfing past the channel). Viewers will typically have a higher average dwell time on networks that carry programs of interest to the viewer.
A method will now be described for profiling viewers on a session-by-session basis and for differentiating viewers in a household of multiple viewers. When the identity (or profile) of a specific viewer or viewers that are watching during any given session can be identified and distinguished from other potential viewers (or profiles), targeted and custom content, including ads, can be delivered to those viewers. Such targeting will be much more efficient and accurate than if the targeting was based solely on household demographics or on a household profile. In a preferred embodiment, the viewer profiling and identification is accomplished by monitoring and processing channel change data, but it could also be accomplished by processing other forms of viewer interactivity including volume adjustments, EPG activity, etc. Moreover, the algorithms and methods described herein are exemplary methods and implementations and are by no means the only possible mechanisms or implementations of the present invention.
The VCPS 290 uses the concept of a viewing session to identify and profile individual viewers and households. In a preferred embodiment, a viewing session is a period of time during which the viewers do not change (i.e., those viewers watching at the start of the session are the same viewers watching at the end of the session). The VCPS 290 identifies the different viewing sessions and the individuals (and their associated profiles) involved in those sessions. Although viewing sessions can be determined and delimited by a viewer's self-identification (e.g., via a viewer specific remote, viewer login or custom menu usage), in the preferred embodiment, automatic detection of viewing sessions and associated viewer or viewers is employed. An exemplary embodiment of automatic detection is now described, but it is to be understood, as will be evident to those skilled in the art that the invention is not limited to the exemplary embodiment described, but could be implemented in a variety of ways.
Viewing sessions can also be defined based on the content accessed by the viewer and the interactivity behavior of the viewer. Examples of such interactivity behavior include but are not limited to, the rate of channel change, program type preferences, preferred networks, and program search patterns. The interactivity behaviors could vary by day part. Using the concept of a sliding window, viewer behavior changes, such as increased channel change frequency, and viewer preference changes, such as changes in programs or networks, can be detected and used to terminate a current session, initiate a new session, or both. A determination as to whether to start a new session based on the change of preferred networks can be made based on predetermined rules regarding the networks (e.g., different networks attract viewers with differing demographic attributes), or could be based on previously stored session information and profiles and/or signature profiles (discussed further herein).
As illustrated on a first time axis 2920A, the sliding window 2900 includes networks DSC, ESPN, and FX. These networks are considered to be identifiable with a single viewer and thus are identified as session 1. As time progresses to time axis 2920B, the sliding window 2900 includes networks FX, ESPN and Lifetime. These changes in networks are not yet enough to trigger a change in session. However, as time progresses to time axis 2920C and the sliding window 2900 includes mostly networks Lifetime and A&E, the VCPS 290 initiates session 2. The determination as to whether to start a new session based on the change of preferred networks can be made based on predetermined rules regarding the networks (e.g., different networks attract viewers with differing demographic attributes), or could be based on previously stored session information and profiles and/or signature profiles (discussed further herein).
This method of automatic session detection and transition is extremely powerful as it is driven by viewer habit and preferences, and occurs in real-time. Although in this example the preferred networks are monitored by the sliding window it is to be understood, as will be evident to those skilled in the art, that other classes of viewer interactivity and preferences or combinations of classes (weighted or non-weighted) could also be used with the sliding window method to identify session transition. These other classes may include, but not limited to, preferred program genre, networks, surf patterns, program categories, viewer type (i.e., male, female, child), volume adjustments, and EPG activity.
In a preferred embodiment, the VCPS 290 uses two sliding windows, a small window to detect radical changes in viewer behavior and a larger window to detect more subtle changes.
The fifteen-minute window 3020 has a high threshold of viewer behavioral change as it is meant to detect radical changes in the viewing behavior in a short period of time. The thirty-minute window 3030 detects more subtle changes in behavior and has a lower threshold for change detection. Each sub-session is labeled with a range of minutes, which indicates the amount of time that has elapsed since session began. Two additional sub-sessions, represent data that is outside the windows 3020, 3030 (events that occurred 30+ minutes ago). One sub-session stores events that occurred over 30 minutes ago for the current day part 3040. The other sub-session stores data that occurred over 30 minutes ago for this session 3050.
As events occur, the data is stored in the 0-4 minute sub-session 3002. After 5 minutes has elapsed, the data for all sub-sessions except the 30+ minute sub-sessions 3040, 3050 is shifted to an adjacent bin (as illustrated, shifted to the right). Data coming out of the 25-29 minute bin 3012 is added to both the 30+ minute day-part bin 3040 and the 30+ minute session bin 3050. The 0-4 minute sub-session 3002 is cleared. The 15-minute sliding window 3020 encompasses the first three sub-sessions 3002-3006 while the 30-minute sliding window 3030 encompasses the first six sub-sessions 3002-3012.
In addition to collecting session data in order to generate a session profile, which is limited to a viewing session and which is collected over a relatively short period of time, the VCPS 290 generates and updates a more complete profile of the user by combining current session data with historical data for this viewer. This aggregated viewer or household historical data is referred to as a signature. A signature can best be described as the aggregation of session data for a particular type of viewer or household over an extended (multiple sessions) period of time. A single signature does not necessarily correspond to a single unique viewer in a household. The combination of multiple members of a household watching television simultaneously may result in a household signature, different than any of the individual signatures.
Also, members of the household with similar viewing habits may be aggregated into a single signature. It is also possible that different viewing habits for a single member of the household can result in multiple signatures for that member. The VCPS 290 stores multiple signatures, in order to account for multiple viewers and subsets of those viewers within a household, each signature representing at least a subset of different viewing characteristics. In a preferred embodiment, the VCPS 290 stores up to fifteen unique signatures per household, and each signature may correspond to a particular viewer, a group of particular viewers, or to the entire household of viewers, and the VCPS 290 retains the signature information in long-term storage (e.g., on a hard drive or in non-volatile memory).
In a preferred embodiment, the VCPS 290 generates and stores viewer and household signature profiles based on one or more session profiles. When the VCPS 290 is first initiated, and there are no session or signature profiles, the VCPS 290 generates the first signature profiles from the first session profiles. Then, as the VCPS 290 monitors and processes data from subsequent viewing sessions, it will continually attempt to correlate the session data with a specific signature, and associate the long term viewing habits and other interactivity of a particular viewer or household with short term viewing habits and interactivity. In an alternative embodiment, signatures may be loaded into the STB 220 externally (either when the unit is manufactured, purchaser, or installed or during use). The loaded signatures may represent characteristics that advertisers are looking for, be standard signatures that identify distinct market segments, be representative signatures for a particular area, or other representations that would still be within the scope of the current invention. The preloaded signatures may be updated to take into account session profiles that are deemed to correlate with the signature profiles or may stay the same until updated by the network operator or other responsible party.
In a preferred embodiment, the signature profile comprises the same profile categories as the session profile allowing for a direct correlation between categories. The VCPS 290 writes session data to the signature history at the end of day parts and sessions. Each 5-minute sub-session tracks the day part that it resides in, and the 30-minute day-part sub-session 3040 does the same. When data is being added from the 25-29 minute sub-session 3012 into the 30+ minute day-part sub-session 3040, the VCPS 290 verifies that the sub-sessions are the same day part 3060. If the sub-sessions are for a different day part, the data from the 30+ minute day-part bin 3040 is written to the signature data that best matches the current session and the 30+ minute day-part bin 3040 is cleared.
Because the VCPS 290 tracks all of the session data and also must write day part data to the signatures, the 30+ minute session sub-session 3050 stores all data for the current session that is outside the sliding windows. This data is used for comparing changes in session behavior. When a session terminates due to behavior change, data within the sliding window represents a new session while data outside that window is the old session.
If the 30 minute window 3030 detected the change in session, the 30+ minute day-part sub-session 3040 is written to the signature file and the 30+ minute day part bin 3040 and 30+ minute session bin 3050 are cleared. Note, the 30+ minute session bin 3050 is not required to be written to the signature file as the 30+ minute day part bin 3040 is written to the signature file, and all previous day parts that were part of this session were previously written to the signature file.
If the 15-minute sliding window 3020 detected the behavior change, then data in the 5-minute sub-sessions outside the 15-minute sliding window 3020 (last 3 five-minute sub-sessions 3008-3012) are also written to the signature file and cleared. When a session terminates due to power off or lack of user interaction, all data in the 5-minute sub-sessions 3002-3012 and the 30+ minute day-part sub-session 3040 are written to the signature file and cleared.
If the new event occurs at the end of a sub-session at step 3103, then a check is made as to whether the leading edge of the sliding window coincides with the end of a day part at step 3117. If they do not coincide, then the last sliding window sub-session is shifted into the 30+ minute session bin at step 3119. If they do coincide, then the day part profile is saved to a signature file at step 3127, the data is cleared from the 30+ minute day part bin at step 3129, and then the last sliding window sub-session is shifted into 30+ minute session bin at step 3119.
At step 3121, a determination is made as to whether there was any change in viewing pattern. If there was no change, all sliding window sub-sessions are shifted at step 3123 and the current bin is cleared at step 3125. The data from the ending sub-session is then processed via the filters at step 3107 and added to the profile for that sub-session at step 3109. If a change in viewing pattern is detected, then the session data is saved to a signature file at step 3131 and a new session is started at step 3133. All sliding window bins are then shifted at step 3133.
As one of ordinary skill in the art would recognize, there are numerous methods for performing the determination of step 3121 that are well within the scope of the current invention. For example, the determination can be made by comparing the current viewing characteristics with the session profile, or, if applicable, with the signature profile associated with the session profile. The current viewing characteristics may be defined as a single element of time or may be multiple elements of time for purposes of comparison. The embodiment described above used two periods of time (last 15 minutes and last 30 minutes) for the current viewing characteristics. It should be noted that the current invention is not limited to these time frames. According to one embodiment, these time frames would be an adaptable parameter of the system.
If multiple time frames are used the thresholds for determining a change in viewership would be different. As described above, the shorter window of time would be used to detect radical changes while the longer window of time would be used to detect more subtle changes. The comparisons may be based on one or more viewing characteristics. If multiple viewing characteristics are used in the comparison, the different characteristics may have different weighting factors applied. If multiple windows of time are used, the same elements may be compared, the same element with different weighting factors may be compared, or different element may be compared.
When the viewing session terminates, the current session profile 3200 is added to the signature profile 3220 that had the best correlation 3210. The VCPS 290 performs a weighted average between the session profile 3200 and the signature profile 3220 so that the data is properly scaled based on overall viewing time. Thus, the signature profiles 3220 that represent long periods of viewing history will only be slightly altered by new viewing data. In addition to adding the session profile 3200 to the associated (highest correlation above threshold) signature profile 3220 at the end of the session, the VCPS 290 may also add the current session profile 3200 to the associated signature profile 3220 at the end of day parts, the end of programs, end of days or defined time periods.
According to one embodiment, once a matching signature has been established, a fully populated 15-minute sliding window 3020 (all three 5 minute bins 3002-3006 are filled) is used to determine viewer behavior changes. The VCPS 290 compares the data in the 15-minute sliding window 3020 against the signature data. The data within the 15-minute sliding window 3020 must differ significantly (e.g., exceed a difference threshold or not meet a correlation threshold) from the signature data for the VCPS 290 to terminate the session. As would be obvious to one of ordinary skill in the art, reducing the difference threshold (or increasing the correlation threshold) would likely result in earlier termination of sessions, and increasing the difference threshold (or decreasing the correlation threshold) would likely result in longer sessions.
In another embodiment, once the matching signature profile has been identified the VCPS 290 may also use a fully populated 30-minute sliding window 3030 to determine viewer changes. As the 30-minute sliding window 3030 is designed to detect more subtle changes, the difference threshold is accordingly lower (or correlation threshold is accordingly higher) than for the 15-minute sliding window 3020. If the 30-minute sliding window 3030 and the signature exceed the difference threshold (or fall below the correlation threshold) the session is terminated. In an alternative embodiment, the 30-minute sliding window 3030 is next compared to the overall session before a termination decision is made. If this difference threshold is also exceeded (or falls below the correlation threshold) the session is terminated. The difference (or correlation) thresholds for both the 30-minute sliding window-vs.-signature and 30-minute sliding window-vs.-overall session may be the same or may be different (to account for the likely differences in the session profile and the signature profile).
The signature summary portion 3380, illustrates a number of signatures identified by their unique ID 3385, total time 3390 and correlation score 3395. The number of signatures displayed may be the top signatures (i.e., top 5), may be all of the signatures stored therein, or may be a predetermined number of signature slots. As illustrated, there are 15 signature slots displayed with the last three being empty signature slots. While it would be possible to store a large amount of signatures on the STB, there is a practical limit to how many should be stored thereon. The limit would be based on the memory in the STB as well as the granularity that is desired in the different signatures. As previously discussed, session monitoring and signature matching preferably occur in real time (or near-real-time), as session data is being received, processed, and continuously compared with signature data. Thus, the correlation scores are updated in real time (or near real time).
The correlation scores 3395 may be based on the correlation of a single element, multiple elements, or multiple weighted elements. According to one embodiment, the correlation score is based on the program genre, surf dwell time and networks. These elements may be weighted with program genre having the heaviest weighting factor. While the viewer type is not used in the correlation score in this embodiment, it may be used to exclude potential matches between a session and a signature. That is, if the session viewer type is significantly different from the signature viewer type, the match will be rejected. If the match is right on the threshold for acceptance, or no match can be found, the VCPS 290 examines the category data to assist in the session to signature match determination. For instance, if the session has a large amount of viewing time in a small set of categories, then the VCPS 290 can search the signature data for high viewing time in those categories.
It is to be noted that the session duration parameter (step 3403) used to determine whether or not the session data should be discarded is not required to be 15 minutes, but could a much smaller or much larger unit of time, or could be an adaptable parameter without departing from the scope of the current invention. Also, the minimum correlation threshold used to determine whether a session and a signature correlate (step 3405), could be set to a relatively high or low value, or could be based on one or more parameters as discussed above without departing from the scope of the current invention. These parameters could be adaptable and set depending on implementation preference. Moreover, as one skilled in the art would recognize there are numerous methods for scoring the signatures (step 3409) that would be well within the scope of the current invention.
According to one embodiment, the VCPS 290 may consider the signatures with the most viewing time to be off limits to purging. If a signature has a large amount of viewing time but the viewer that created it is no longer with the household, the signature will eventually be able to be deleted after the system accumulates viewing time for the active signatures. For the signatures that can be deleted, a combination of total viewing time and date that the signature was last updated contribute to the signature that the system selects for purging. A STB box has N slots for signatures and only signatures in the top N/2 slots can be considered off limits for purging. The system then calculates the total viewing time T for all signatures. If a signature is in a top slot and it has a viewing time of at least T/2N, then the signature cannot be deleted.
For example, assume that the STB has 15 signatures (N=15) stored therein and has compiled 50,000 seconds of total viewing time for all signatures (T=50,000). A signature cannot be purged if it is within the top 7 signatures (15/2) of viewing time and the signature viewing time is at least 1,667 seconds (50000/30).
The signatures that can be deleted will be assigned a score based on the viewing time and date of last update. Given a viewing time of V and number of days since last update of D, then the score will be V/((D/10)+1). Thus basically states that every 10 days of signature inactivity reduces the value of the viewing time. For example, if the viewing time is 2000 seconds and it has been 12 days since the last update, the score will be 1000 [2000/((12/10)+1)].
Note that the calculation uses integer math, which rounds the result down. Thus, the denominator includes a +1 to ensure that the value will not be 0.
The following example demonstrates all of these concepts. This example has six signature bins and a total viewing time of 60,000 seconds. Thus, the top three signatures that have a minimum signature time of 5000 seconds (60000/12) may be marked as off limits. However, as illustrated only two of the bins meet the viewing time and are marked off limits. Out of the remaining four signatures, signature 5 has the lowest score of 2000 [40004(16/10)+1)].
After the event (power on, channel change, end of program) is processed at step 3517, the profile characteristics associated with the event are added to the current session profile (step 3519). In accordance with a preferred embodiment of the current invention, the characteristics of the event are added on a time-weighted basis. Based on the current channel that is tuned to, the VCPS 290 determines when the current program is scheduled to end (based on program data) and adds an event for the end of the program to the event queue (step 3521).
If the event was determined not to be an end of program event, the VCPS determines if the event was an end of bin event as established in step 3509 (step 3523). If the event was an end of bin event, the end of bin event is processed (step 3525) and the characteristics of this event are added to the current session profile (step 3527). Next a determination is made as to whether it is the end of a day part (step 3531). If it is determined that it was the end of a day part, the current session profile for that day part (i.e., bin 3040) is saved to the associated signature profile (step 3531) and then cleared (step 3533). Regardless of whether, the event was an end of day part or not, the data within the sliding windows is shifted to the next bin (step 3535). Referring back to
If the event was not an end of pin event, the VCPS 290 determines if the event was a power off event (step 3539). If the event was not a power off event, the process returns to step 3505. If the event was a power off event, the power off event is processed (step 3541), the characteristics of this event are added to the current session profile (step 3543), the session profile is added to the associated signature profile, preferably on a time weighted basis (step 3545), and the event queue is cleared (step 3547). The process is complete at step 3549.
As one of ordinary skill in the art would recognize, this process is merely an exemplary process. The steps of the process could be rearranged, steps could be removed or added, or completely different steps could be used to accomplish the same or similar result without departing from the scope of the current invention.
According to one embodiment, the VCPS 290 also periodically compares each signature with every other signature, in order to identify signatures that may correlate with one another. For example, as session data is added to existing signatures, it is possible that some of the signatures will begin to correlate with one another, indicating that the signatures actually represent the same viewer or viewer type. In one embodiment, the VCPS 290 handles this case by attempting to correlate signatures after every signature update. These updates preferably occur at the end of day parts and the end of sessions. The process of merging or joining two correlated signatures is accomplished by comparing the recently updated signature with every other signature stored by the VCPS 290. The signature that best matches (according to a predefined threshold) will be merged with the recently updated signature by calculating the weighted average of the signature data. Additionally, the VCPS 290 preferably performs the process iteratively, attempting to merge the new signature with all other signatures. This iterative procedure is carried out because combining the two signatures may have altered them enough to correlate with another signature.
According to one embodiment, the VCPS 290 compares most of the elements of the signature data in order to determine if signatures correlate with one another. Unlike the session to signature comparisons, category data is used in the signature correlation process because the signatures represent many sessions, and the correlation of category data is particularly useful when comparing large data sets. However, viewer type is not used to correlate signatures, it is only used to exclude matches. That is, if the VCPS 290 determines that two signatures correlate, the VCPS 290 will look at viewer type. As long as the viewer type is not significantly different, the VCPS 290 will merge the two correlated signatures. However, if the viewer type is significantly different (i.e., signature 1=0.8 male, 0.1 female, 0.1 child and signature 2=−0.1 male, 0.8 female, 0.1 child) then the VCPS 290 will not merge the two signatures.
The VCPS 290 can perform the comparisons either after updating a signature with session data or as a background task. Performing the comparisons after an update will require only comparing the updated signature with the other signatures. Background comparisons must compare all signatures against all others. If two or more signatures correlate beyond a certain correlation threshold, the VCPS 290 combines signatures by performing a weighted average of the data from each signature. Such a correlation threshold may be dynamically configurable (e.g., from instructions from the HE).
In an alternate embodiment, standardized or customized signatures are downloaded from the HE 210 and stored at the STB 220 for utilization by the VCPS 290. Such custom signatures may be derived based on, for instance, expected or predicted viewer behavior, or alternatively, these signatures may represent aggregate session data from multiple households. An example of such a signature is one that reflects one or more demographic traits shared by a cluster or group of households, which may be identified by analyzing session profile data (or signatures) from an array of households, and grouping households together that have highly correlative session or signature profiles. Session profiles can then be correlated with these “standard” signatures (e.g., a form of collaborative filtering) as well as with locally generated signatures as described above.
As discussed above, the VCPS 290 correlates profiles with other profiles, including, but not limited to:
There are numerous ways that profiles can be correlated with each other. The correlation can range from simple to complex. The complexity of the correlation depends on such factors as what is being correlated, and how much data is contained within the profiles. For example, a correlation of a session profile against itself may be quite simple. That is, a single profile (i.e., network profile) may be monitored to determine when a session ends. On the other hand, a correlation of a session profile to a signature profile may be more complex and may correlate numerous profiles (i.e., genre, networks, category, surf) to determine when there is a match. Some possible profile correlations and the value of each are discussed below.
Genre profiles indicate the amount of time that a viewer watches each type of program genre, in effect weighting the importance of that program type to the viewer. The sessions and signatures track the viewing times of each genre in an array. The correlation compares the elements of each of the genre arrays. As it is believed that viewers generally watch the same types of programs, using the genre in the correlation calculation provides a highly accurate method for quickly comparing two profiles.
Channel surf dwell time profiles capture how long a particular channel or network will maintain the viewers' interest during a surf routine. Because a channel surf sequence may only take one or two minutes, a major benefit of tracking channel surf behavior is that the dwell time data is collected over a short period of time when compared with viewing interests. The sessions and signatures track dwell times by network, and each one has an array of both the number of times a network is visited during the surf sequence and the total dwell time on each network, in order to calculate an average dwell time per network per channel surf sequence. The average dwell time per network is correlated.
Network profiles track the amount of time that the viewer watches a particular network by day part and day of week. Because a viewer will typically watch the same programs at the same time every week, correlating the data by the appropriate day and day of week results in matching a viewer session to the appropriate signature. In one embodiment, the session-to-signature correlation compares the network information for the current day and day part against the signature data for the same time. The signature-to-signature correlation used to merge signatures compares all network day and day part data. The average correlation is used to help determine whether the signatures match. Since the signatures aggregate data over many sessions, the network viewing by day part should be similar if the signatures actually represent the same viewer.
Program category profiles are much more specific than genres and can be utilized, for instance, if a session cannot be matched with a signature or only produces marginal matches, to match a session with a signature. For example, if the VCPS 290 collects data reflecting a large amount of viewing time in a small set of categories, the VCPS 290 can search the history for signatures with large viewing times in those same categories. The search results can be used to determine if a marginal match between session and signature is actually a match. It can also help classify a session that would not otherwise correlate with a signature.
As would be obvious to one of ordinary skill in the art, the amount of data gathered for each session and each signature is dependent upon the amount of memory in the STB 220, the amount of granularity (and potentially accuracy) desired. A tradeoff needs to be made between the amount of memory required and the granularity/accuracy desired. According to one embodiment, each session profile includes the following category profiles:
According to one embodiment, each signature profile includes the following category profiles:
The VCPS 290 may track prime time viewership (8 PM-11 PM) for the weekdays, evening viewing (6 PM-8 PM) for all weekdays, late night viewing (11 PM-5 AM) for all weekdays, daytime viewing (5 AM-6 PM) for all weekdays and weekend viewership. According to this embodiment, when comparing session data to signature data, the VCPS 290 compares the following data as the primary match criteria: session genre data and signature genre data; session network data by day part with signature network data by same corresponding day part; and session surf dwell time with signature surf dwell time. Each element or category may be weighted based on the importance of the data in the correlation. Genre has the largest weighting factor in this embodiment, but other relative weightings are also possible. The correlation result is a score for the signature match. The signature that has the highest correlation with the session is considered the best match, and the session data may be associated with that signature.
The signature profiles may be used to target content (advertisements, pay per view (PPV) events, video on demand (VOD) programming) to the subscribers or to customize their viewing environment (i.e., favorite programs listed first in EPG, format of EPG). The signature profiles stored in the STB 220 may be transmitted to the HE at determined intervals (i.e., every night at off-peak hours, every week). The HE may aggregate the data received from each of the STBs 220 connected thereto in order to form groups of viewers with similar characteristics. The groups having similar characteristics may receive specific content targeted to their characteristics (signature profiles). According to another embodiment, the HE may also use other external data to help group subscribers. The external data may be in the form of geodemographic data, such as the demographic data associated with distinct ZIP+4 geographies provided by Claritas (see U.S. Provisional Application No. 60/238,059; U.S. patent application Ser. No. 09/635,542; and U.S. patent application Ser. No. 09/928,024, each of which were previously referred to and incorporated by reference for additional details). The external data may also be in the form of other transactional profiles, such as purchase transactions (see U.S. patent application Ser. No. 09/591,577; U.S. patent application Ser. No. 09/268,519; and U.S. patent application Ser. No. 09/928,024, each of which were previously referred to and incorporated by reference for additional details).
The groups may be formed based on the topography of the television viewing network (i.e., based on nodes within the system). The groups would then consist of clusters of subscribers, wherein each of the clusters could be associated with a node, branch or other element of the system (see U.S. Provisional Application No. 60/278,612 and U.S. patent application Ser. No. 09/928,024, each of which were previously referred to and incorporated by reference for additional details). In an alternative embodiment, the STB may receive the additional data (i.e., geodemographic data, other transaction data) directly and incorporate this data into the generated profiles (i.e., signature profiles) to develop enhanced profiles. These enhanced profiles can then be forwarded to the HE for aggregation and potential grouping as described above.
Regardless of where the enhanced profiles are generated, there is likely a practical limit to how many different groups are formed (i.e., 5). The groups may be formed and the advertisements targeted thereto, or ad profiles defining characteristics about the target market of the ad are created and the users are grouped to these profiles. As discussed above, each of the groups may receive material targeted for them. In one embodiment, everybody will receive all content and will select the appropriate content for display based on matching the group number associated with the content and the group number of the STB, or particular user or group of users interacting with the STB. In a preferred embodiment, each STB will receive only the content associated with it based on what group it falls in, thus saving bandwidth. If the content is targeted ads, the targeted ads may be inserted in place of the default ads to create a plurality of presentation streams (programs with targeted ads). Local cable companies have the equipment necessary to insert targeted ads as they are permitted to substitute local ads in place approximately 20% of the default ads. In the preferred embodiment, it is necessary for the television system to be able to route the different presentation streams to different areas (i.e., nodes, branches) within the system so that each user only receives the appropriate presentation stream (see U.S. patent application Ser. No. 09/553,637; International Patent Application No. PCT/US00/10633; U.S. patent application Ser. No. 09/553,099; U.S. patent application Ser. No. 09/694,848; International Patent Application No. PCT/US00/29034; and U.S. patent application Ser. No. 09/750,800, each of which were previously referred to and incorporated by reference for additional details).
In an alternative embodiment, the STB will insert specific content (i.e., targeted ads) locally. Thus, each subscriber may, in effect, have ads targeted directly to them. The ads may be inserted in avails (advertisement opportunities) regardless of the programming being viewed. The STB may receive the targeted ads via an ad channel. The ads may be delivered ahead of time and stored on the STB, or be received at approximately the same time as the avail and inserted on the fly. If the ads are to be stored on the STB, the ad channel may deliver a plurality of ads on the ad channel with the STB selecting the appropriate ads to store thereon. The selection may be made based on group designations or may be made by comparing the ad profile with the signature profiles on the STB. In addition to the ads, an ad queue will need to be generated and stored on the STB to define what order the ads are inserted and what criteria affect the criteria (see U.S. patent application Ser. No. 09/712,790; International Patent Application No. PCT/US01/27217; U.S. patent application Ser. No. 09/731,606; U.S. patent application Ser. No. 09/748,943; U.S. patent application Ser. No. 09/750,800; and U.S. patent application Ser. No. 09/824,434, each of which were previously referred to and incorporated by reference for additional details).
In an alternative embodiment, the ad queue may be stored on the STB while the ads are stored somewhere else within the network. When the ad queue determines the next ad to be inserted, the ad is retrieved from the network and delivered to the STB. The ad may be delivered over an ad channel or over a dedicated line (such as an Internet connection). If the ad channel was used the ads may be delivered in real time (would require coordination between program(s) and ad channel), or ahead of time and temporarily stored (i.e., start delivering one avail in advance). The ads could be delivered on the ad channel either at the actual bit rate for the ad or at a slower bit rate where portions of the ad are stored as they are received. As one skilled in the art would recognize, the Internet connection provides both benefits and drawbacks. The drawback is the need the separate connection. The benefit is that the queue points to the URL address for the ad so the ad is displayed upon selection from the queue (see U.S. patent application Ser. No. 09/742,506, which was previously referred to and incorporated by reference for additional detail).
In a preferred embodiment, the ad queue is actually a plurality of ad queues with a different queue associated with each signature, or group of signatures. When the VCPS 290 determines that a session has ended and thus a new user or group of users (identified by a correlation with a new signature) is interacting with the TV, the ad queue changes to the one associated with that signature. Moreover, the ad queue may be adjusted based on the programs being watched as certain advertisers may not want ads displayed during certain programs or may pay a premium for insertion in other programs (See U.S. patent application Ser. No. 09/742,852, which was previously referred to and incorporated by reference for additional detail).
According to another embodiment, the STB may also store ads and ad queues for other types of targeted advertisements such as EPGs, program bugs (overlay in the corner of the video), product placement (illustration of product placed in program, such as a Coke can placed in actors hand), trick play ads (shortened version of ad displayed when user fast forwards through recorded ad), and record ads (alternative ads placed in a recorded program if it is determined that different subscriber is viewing or if it is being viewed for more than the first time). The ad queue could be used to manage how all of these ads are displayed so as not to saturate the viewer but yet enhance the advertisement effectiveness. The various types of ads and possible coordination therebetween are discussed in U.S. Provisional Patent Application No. 60/238,056; U.S. patent application Ser. No. 09/749,255; International Patent Application No. PCT/US01/31682; U.S. patent application Ser. No. 09/680,622; U.S. Provisional Patent Application No. 60/281,067; and U.S. Provisional Patent Application No. 60/329,992, each of which have all previously been referred to and incorporated by reference for additional detail.
Regardless of what is being correlated (ad profile to signature profile, session profile to signature profile), how many different elements are being correlated (program genre; program genre, channel changes and networks), and whether the different elements are weighted or not, there are numerous ways to perform these correlations. If the profiles are in the form of vectors you can perform a scalar dot product on the vectors to determine the correlation between the two vectors. In order for a scalar dot product to work the vectors must be normalized. That is, the magnitude of the vector must be 1.0. To calculate the magnitude, take the square root of the sum of the squares of the components as shown below.
To normalize the vector, divide each component by the magnitude of the vector. Sample dot products are shown below.
One problem with the simple scalar dot product is that, it is possible to have vectors that do not match each other have the same scalar dot product as vectors that are identical (first illustrated vector above).
According to another embodiment, the variance (σ2) and standard deviation (σ) of vectors may be used to calculate the correlation. Assuming the two vectors that we wish to correlate are labeled A and B, the variance is defined by the following function:
The equation produces a low variance if there are only small differences between the components of each vector. This indicates a high correlation. A variance of zero indicates that there is no difference between the two vectors. Since each distribution vector is normalized, the square of the difference of each component will be in the range of 0 through 1. Since the variance is the average of this value, the variance will be limited to the same range. The actual variances will be small since the differences between components of a distribution will be small, and the differences are squared.
The standard deviation, σ, is the square root of the variance. Since the variance will be very small, the square root will increase the value. Therefore, it is suggested that the standard deviation be used as the basis of the correlation value. As a correlation of 1.0 would reflect a perfect match a standard deviation of 0.0 is a perfect match. Since the standard deviation will be constrained to values between 0 and 1, the following equation is suggested for the correlation:
correlation=1−σ
Using the above examples, the correlations are now calculated as:
The above examples of correlation methods are exemplary in nature and in no way are intended to limit the scope of the invention. As would be obvious to one or ordinary skill in the art, there are numerous other methods for determining the correlation between profiles that would be well within the scope of the current invention.
As will be evident to those of ordinary skill in the art, the VCPS 290 application and other software components, can be implemented in a variety of software languages, including C, C++, and Java. Moreover, the VCPS can be implemented on and/or integrated with a variety of STB platforms, including those boxes produced by Motorola (formerly General Instrument) and Scientific Atlanta, and with a variety of operating systems, including VxWorks and PowerTV, and with a variety of middlewares such as those produced by Liberate and OpenTV.
Although this invention has been illustrated by reference to specific embodiments, it will be apparent to those skilled in the art that various changes and modifications may be made which clearly fall within the scope of the invention. The invention is intended to be protected broadly within the spirit and scope of the appended claims.
This application is a continuation of U.S. patent application Ser. No. 14/196,488 (now U.S. Pat. No. 9,473,814), filed Mar. 4, 2014, which is a continuation of U.S. patent application Ser. No. 13/277,839 (now U.S. Pat. No. 8,667,536), filed Oct. 20, 2011, entitled Profiling and Identification of Television Viewers, which is a continuation of U.S. patent application Ser. No. 11/751,154 (now U.S. Pat. No. 8,046,798), filed May 21, 2007, entitled Profiling and Identification of Television Viewers, which is a divisional of U.S. patent application Ser. No. 09/998,979 (now U.S. Pat. No. 7,260,823), filed Oct. 31, 2001, and entitled Profiling and Identification of Television Viewers, the entire disclosures of which are incorporated herein by reference. U.S. patent application Ser. No. 09/998,979 (now U.S. Pat. No. 7,260,823) claims the benefit of U.S. Provisional Application Nos. 60/260,946, filed Jan. 11, 2001, entitled Viewer Profiling within a Set-top Box, and 60/263,095, filed Jan. 19, 2001, entitled Session Based Profiling in a Television Viewing Environment, the entire disclosures of which are incorporated herein by reference. This application is related to all of the Applicants' applications, patents and publications listed below. The entire list below is herein incorporated in their entirety, but are not admitted to be prior art: U.S. patent application Ser. No. 09/204,888, filed on Dec. 3, 1998, now U.S. Pat. No. 7,150,030 entitled “Subscriber Characterization System”;PCT Application PCT/US99/28528 (WIPO Publication WO 00/33160A3), filed on Dec. 2, 1999 entitled “Subscriber Characterization and Advertisement Monitoring System”;U.S. patent application Ser. No. 09/516,983 now U.S. Pat. No. 7,240,355, filed on Mar. 1, 2000, entitled “Subscriber Characterization with Filters”;U.S. patent application Ser. No. 09/591,577, now U.S. Pat. No. 7,949,565, filed on Jun. 9, 2000, entitled “Privacy-Protected Advertising System”;U.S. patent application Ser. No. 09/635,252, filed on Aug. 9, 2000, now abandoned, entitled “Subscriber Characterization Based on Electronic Program Guide Data”;U.S. patent application Ser. No. 09/205,653, filed on Dec. 3, 1998, now U.S. Pat. No. 6,457,010, entitled “Client-Server Based Subscriber Characterization System”;U.S. patent application Ser. No. 09/205,119, filed on Dec. 3, 1998, now abandoned, entitled “Advertisement Monitoring System”;U.S. patent application Ser. No. 09/516,314, filed on Mar. 1, 2000, now abandoned, entitled “Advertisement Monitoring and Feedback System”;U.S. patent application Ser. No. 09/452,893, filed on Dec. 2, 1999, now U.S. Pat. No. 6,684,194, entitled “Subscriber Identification System”;PCT Application PCT/US99/28600 (WIPO Publication WO 00/33233), filed on Dec. 2, 1999 entitled “Subscriber Identification System”;U.S. patent application Ser. No. 09/635,253, filed on Aug. 9, 2000, now U.S. Pat. No. 6,714,917, entitled “Subscriber Identification Based on Electronic Program Guide Data”;U.S. patent application Ser. No. 09/268,519, filed Mar. 12, 1999, now U.S. Pat. No. 6,298,348, entitled “Consumer Profiling System”;PCT Application PCT/US99/28628 (WIPO Publication WO 00/33228), filed on Dec. 2, 1999, entitled “Consumer Profiling and Advertisement Selection System”;U.S. patent application Ser. No. 09/553,637, filed on Apr. 20, 2000, now U.S. Pat. No. 6,820,277, entitled “Advertising Management System for Digital Video Streams”;PCT Application PCT/US00/10633 (WIPO Publication WO 00/64165), filed on Apr. 20, 2000, entitled “Advertising Management System for Digital Video Streams”;U.S. patent application Ser. No. 09/553,099, filed on Apr. 20, 2000, now U.S. Pat. No. 6,820,277, entitled “Advertisement Insertion Techniques for Digital Video Streams”;U.S. patent application Ser. No. 09/568,084, filed on May 10, 2000, now abandoned, entitled “Advertisement Subgroups for Digital Video Streams”;PCT Application PCT/US00/12710 (WIPO Publication WO 00/69163), filed on May 10, 2000 entitled “Advertisement Subgroups for Digital Video Streams”;U.S. patent application Ser. No. 09/568,477, filed on May 10, 2000, now U.S. Pat. No. 6,615,039, entitled “Advertisement Subgroups for Digital Video Streams”;U.S. patent application Ser. No. 09/635,539, filed on Aug. 10, 2000, now abandoned, entitled “Delivering targeted advertisements in cable-based networks”;U.S. patent application Ser. No. 09/694,848, filed on Oct. 20, 2000, now U.S. Pat. No. 7,068,724, entitled “Method and Apparatus for Inserting Digital Media Advertisements into Statistical Multiplexed Streams”;PCT Application PCT/US00/29034 (WIPO Publication WO 01/30086A), filed on Oct. 20, 2000, entitled “Method and Apparatus for Inserting Digital Media Advertisements into Statistical Multiplexed Streams”;U.S. patent application Ser. No. 09/796,339, filed on Feb. 28, 2001, now abandoned, entitled “Privacy-Protected Targeting System”;PCT Application PCT/US01/06650 (WIPO Publication WO 01/65453A1), filed on Feb. 28, 2001 entitled “Privacy-Protected Targeting System”;U.S. patent application Ser. No. 09/635,542, filed on Aug. 10, 2000, now abandoned, entitled “Grouping Subscribers Based on Demographic Data”;U.S. patent application Ser. No. 09/635,544 filed on Aug. 10, 2000, now abandoned, entitled “Transporting Ad Characterization Vectors”;U.S. patent application Ser. No. 09/712,790 filed on Nov. 14, 2000, now U.S. Pat. No. 7,068,724, entitled “Queue Based Advertisement Scheduling and Sales”;PCT Application PCT/US01/27217, filed on Aug. 31, 2001, entitled “Targeted Advertising at the Set-Top Box”;U.S. patent application Ser. No. 09/731,606 filed on Dec. 7, 2000, now abandoned, entitled “System for Providing Targeted Advertisements Using Advertiser Specific Target Groups”;U.S. patent application Ser. No. 09/748,949 filed on Dec. 27, 2000, now abandoned, entitled “Advertisement Filtering and Storage for Targeted Advertisement Systems”;U.S. patent application Ser. No. 09/742,527 filed on Dec. 21, 2000, entitled “System and Method for Automatically Managing Avail Inventory Data and Avail Pricing”;U.S. patent application Ser. No. 09/748,942 filed on Dec. 27, 2000, now U.S. Pat. No. 7,328,448, entitled “Advertisement Distribution System for Distributing Targeted Advertisements in Television Systems”;U.S. patent application Ser. No. 09/742,506 filed on Dec. 21, 2000, now abandoned, entitled “Internet Based Electronic Program Guide Advertisement Insertion Method and Apparatus”;U.S. patent application Ser. No. 09/748,943 filed on Dec. 27, 2000, now abandoned, entitled “Delivering Targeted Advertisements to the Set-Top Box”;U.S. patent application Ser. No. 09/742,534 filed on Dec. 21, 2000, now U.S. Pat. No. 7,039,932, entitled “Queue Based Head-End Advertisement Scheduling Method and Apparatus”;U.S. patent application Ser. No. 09/742,852 filed on Dec. 21, 2000, entitled “System for Rescheduling and Inserting Advertisements”;U.S. patent application Ser. No. 09/750,800 filed on Dec. 28, 2000, now U.S. Pat. No. 7,228,555, entitled “System and Method for Delivering Targeted Advertisements Using Multiple Presentation Streams”;U.S. patent application Ser. No. 09/766,004 filed on Jan. 19, 2001, now U.S. Pat. No. 7,185,353, entitled “System and Method for Delivering Statistically Scheduled Advertisements”;U.S. patent application Ser. No. 09/824,434 filed on Apr. 2, 2001, now abandoned, entitled “Grouping of Advertisements on an Advertising Channel in a Targeted Advertising System”;U.S. patent application Ser. No. 09/658,204 filed on Sep. 8, 2000, now abandoned, entitled “Targeted Advertising Through the Electronic Program Guide”;U.S. Provisional Application No. 60/238,059, filed on Oct. 5, 2000, entitled “Platform Independent Addressable Television Advertising System”;U.S. Provisional Application No. 60/238,056, filed on Oct. 5, 2000, entitled “Method and System for Addressable Advertising in the Electronic Program Guide”;U.S. patent application Ser. No. 09/749,255 filed on Dec. 27, 2000, now U.S. Pat. No. 7,653,923, entitled “Scheduling and Linking IPG Ads in Conjunction with Programming Ads in a Television Environment”;PCT Application PCT/US01/31682, filed on Oct. 5, 2001, entitled “Targeting Ads on IPG's Live Programming and Recorded Programming, and Coordinating the Ads there between”;U.S. patent application Ser. No. 09/680,622 filed on Oct. 6, 2000, now abandoned, entitled “Method and System for Addressable and Program Independent Advertising During Recorded Programs”;U.S. patent application Ser. No. 09/751,349 filed on Dec. 28, 2000, now abandoned, entitled “Inserting Local Signals During MPEG Channel Changes”;U.S. Provisional Application No. 60/267,370, filed on Feb. 8, 2001, entitled “Presentation Stream Switching Using Channel Maps”;U.S. Provisional Application No. 60/278,612, filed on Mar. 26, 2001, entitled “Formation and Utilization of Cable Microzones”;U.S. Provisional Application No. 60/281,037, filed on Apr. 3, 2001, entitled “Personal Video Recorder (PVR) Market Overview and Advertising Opportunities”;U.S. Provisional Application No. 60/329,992, filed on Oct. 17, 2001, entitled “Personal Video Recorder (PVR) System Requirements and Specification”;U.S. patent application Ser. No. 09/928,024 on Aug. 10, 2001, entitled “Targeting Ads to Subscribers based on Privacy Protected Subscriber Profiles”; andPCT Application PCT/US01/25261, filed on Aug. 10, 2001, entitled “Targeting Ads to Subscribers based on Privacy Protected Subscriber Profiles”.
| Number | Name | Date | Kind |
|---|---|---|---|
| 4258386 | Cheung | Mar 1981 | A |
| 4546382 | McKenna | Oct 1985 | A |
| 4602279 | Freeman | Jul 1986 | A |
| 4745549 | Hashimoto | May 1988 | A |
| 5251324 | McMullan, Jr. | Oct 1993 | A |
| 5374951 | Welsh | Dec 1994 | A |
| 5550928 | Lu | Aug 1996 | A |
| 5600364 | Hendricks | Feb 1997 | A |
| 5608445 | Mischler | Mar 1997 | A |
| 5848396 | Gerace | Dec 1998 | A |
| 5857190 | Brown | Jan 1999 | A |
| 5948061 | Merriman | Sep 1999 | A |
| 5991735 | Gerace | Nov 1999 | A |
| 6005597 | Barrett | Dec 1999 | A |
| 6067440 | Diefes | May 2000 | A |
| 6119098 | Guyot | Sep 2000 | A |
| 6177931 | Alexander | Jan 2001 | B1 |
| 6201536 | Hendricks et al. | Mar 2001 | B1 |
| 6236978 | Tuzhilin | May 2001 | B1 |
| 6286140 | Ivanyi | Sep 2001 | B1 |
| 6298348 | Eldering | Oct 2001 | B1 |
| 6446261 | Rosser | Sep 2002 | B1 |
| 6449632 | David | Sep 2002 | B1 |
| 6457010 | Eldering | Sep 2002 | B1 |
| 6463585 | Hendricks | Oct 2002 | B1 |
| 6615039 | Eldering | Sep 2003 | B1 |
| 6684194 | Eldering | Jan 2004 | B1 |
| 6684197 | Kolls | Jan 2004 | B1 |
| 6714917 | Eldering | Mar 2004 | B1 |
| 6820277 | Eldering et al. | Nov 2004 | B1 |
| 7039932 | Eldering | May 2006 | B2 |
| 7068724 | Hamilton | Jun 2006 | B1 |
| 7150030 | Eldering et al. | Dec 2006 | B1 |
| 7185353 | Schlack | Feb 2007 | B2 |
| 7228555 | Schlack | Jun 2007 | B2 |
| 7240355 | Eldering et al. | Jul 2007 | B1 |
| 7328448 | Eldering et al. | Feb 2008 | B2 |
| 7653923 | Flickinger | Jan 2010 | B2 |
| 7949565 | Eldering et al. | May 2011 | B1 |
| 8046798 | Schlack et al. | Oct 2011 | B1 |
| 8180675 | Blasko et al. | May 2012 | B2 |
| 20010049620 | Blasko | Dec 2001 | A1 |
| 20020026638 | Eldering et al. | Feb 2002 | A1 |
| 20020072966 | Eldering et al. | Jun 2002 | A1 |
| 20020083439 | Eldering | Jun 2002 | A1 |
| 20020083441 | Flickinger et al. | Jun 2002 | A1 |
| 20020083445 | Flickinger et al. | Jun 2002 | A1 |
| 20020087973 | Hamilton et al. | Jul 2002 | A1 |
| 20020095676 | Knee | Jul 2002 | A1 |
| 20020123928 | Eldering | Sep 2002 | A1 |
| 20020144263 | Eldering et al. | Oct 2002 | A1 |
| 20030101449 | Bentolila | May 2003 | A1 |
| 20060195857 | Wheeler | Aug 2006 | A1 |
| 20060248558 | Barton | Nov 2006 | A1 |
| Number | Date | Country |
|---|---|---|
| 9741673 | Nov 1997 | WO |
| 9828906 | Jul 1998 | WO |
| 9960789 | Nov 1999 | WO |
| 0022818 | Apr 2000 | WO |
| 0033160 | Jun 2000 | WO |
| 0033224 | Jun 2000 | WO |
| 0033228 | Jun 2000 | WO |
| 0033233 | Jun 2000 | WO |
| 0049801 | Aug 2000 | WO |
| 0054504 | Sep 2000 | WO |
| 0064165 | Oct 2000 | WO |
| 0069163 | Nov 2000 | WO |
| 0130086 | Apr 2001 | WO |
| 0165453 | Sep 2001 | WO |
| 0213112 | Feb 2002 | WO |
| 0230112 | Feb 2002 | WO |
| 0219581 | Mar 2002 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 60260946 | Jan 2001 | US | |
| 60263095 | Jan 2001 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 09998979 | Oct 2001 | US |
| Child | 11751154 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 14196488 | Mar 2014 | US |
| Child | 15296793 | US | |
| Parent | 13277839 | Oct 2011 | US |
| Child | 14196488 | US | |
| Parent | 11751154 | May 2007 | US |
| Child | 13277839 | US |
