Patent Application
20050120368
The present invention relates to a programme suggestion engine which automatically selects a subset of video signals from a source set of video signals containing recorded TV programmes. In particular, the present invention relates to an automatic programme summary system as a part of the respective programme suggestion engine for automatically creating a summary of a programme forming the content of the source set of video signals based on meta data available for the content.
With the growing number of video channels, it becomes very difficult for a consumer to select and view only those programmes in which he is really interested. Furthermore, the interesting programs are frequently broadcasted at the wrong time, e.g. when the consumer is asleep or at work. Sometimes, two interesting programmes are broadcasted at the same time so that a consumer has to decide in which programme he is interested most. With the possibility of taping programmes on a video recorder, this decision can be altered into which of the two competing programmes he or she is momentarily interested in most, e.g. the possibly important soccer match or the possibly thrilling interesting feature film.
Although a video recorder offers a consumer the liberty to enjoy the programmes of interest at a convenient time, he or she has still to be informed in advance on what will be presented on the various TV channels that can be received. The consumer therefore frequently has to consult a TV programme guide to find out what will be broadcasted at what time. If the programme guide is in printed form, the consumer further has to program the video recorder manually.
In recent years, electronic programme guides (EPG) have been developed as an alternative to print programme guides. An EPG is an application used with appliances for the reception of television programmes that lists current and scheduled programmes which will be available on a multitude of TV channels. An EPG may also be accessed via the Internet. It provides a short summary and a comment for each programme and can be used to program e.g. a video cassette recorder (VCR). An EPG can be regarded as the electronic equivalent of a printed television programme guide.
An EPG is usually accessed with a remote control. It provides a menu for the consumer to view a list of programmes scheduled for the next few hours up to a few of the next days. For the protection of children and young people, parental controls are typically incorporated in the EPG. An EPG typically allows order-per-view programming, search for programmes based on a certain genre and to set a VCR for recording selected programmes. As currently each digital television provider offers its own consumer interface and content for the EPG, the format of the EPG is so far highly dependent on the respective provider. General standards like e.g. the NHP-standard are not yet in force.
One drawback of the printed as well as the electronic television programme guide is, that a consumer still has to look through the programme offers prior to the starting time of the respective programmes. But many times, a consumer does not like to be bothered to go through all the offers presented in the programme guide and prefers to look for an interesting programme just at the time he or she got ready to watch TV, i.e. independent of the broadcasting time of a possibly interesting programme. A consumer is hereby supported by a TV programme suggestion engine which records all TV programmes on the various TV channels which might be of interest to the consumer and prepares a recommendation for the consumer on the basis of the recorded material when requested. The consumer then only has to select those programmes from the recommendation which interest him or her most at that particular moment.
Although a consumer might have a special interest in a certain genre of programmes, it might still be difficult for her or him to decide if she or he is really interested in just the special programme offered, particularly when the actual content of the programme is not known to him like is typically the case with recently taped soccer games or with feature films broadcasted for the first time or something similar like that. An EPG usually will provide some comment on a programme but one cannot rely on it, and sometimes the respective comment may also not be meaningful to the consumer. Due to their nature, taped live sport events are usually not provided with a comment. A consumer is therefore many times forced to view a programme before finding out, that it was not what he would have liked to see at the very special moment.
It is therefore desirable to provide a consumer with some information about important events present in a programme of possible interest. To this respect, a key event detector is used in a programme suggestion engine for verifying the occurrence of particular events in a programme. Usually, the set of video signals recorded for the consumer comprises more than one recommended programme. The recording of the programmes is selected on the basis of information about preferences of the consumer for certain types of programmes. The respective type of programmes is typically defined by a generic term like e.g. sports, feature movies, music, news, shows or the like. But also more detailed genre distinction may be used like e.g. soccer, motor sport or tennis. The recorded source set of video signals is then subdivided into sub-sections, each representing a programme of a particular genre.
The key event detector examines a sub-section for identifying therein all segments corresponding to pre-defined events like e.g. in case of a soccer match corresponding to a goal or a penalty. The key event detector hereto processes the audio channel of the source set of video signals and creates meta data characterising particular signal sequences in the source set of video signals. One typical information available via the meta data is e.g. the generic term of the programme contained in the respective sub-section the meta data refer to.
For identifying the events occurring in a programme, the key event detector first examines the audio signal of the related sub-section for known characteristics allowing a characterisation of the possible origin of certain sections of the audio signal as for instance the section representing speech, music, noise, silence, mood or gender of a speaker or the like. Each such characterised coherent sequence of audio signal forms a sub-unit within the sub-section. Each of the sub-sections is then classified according to its determined type character, and the key event detector creates meta data corresponding to the respective classifications.
The incident represented by a classified sub-unit is usually too isolated to allow at least an idea about the related programme scene being of possible interest or not. A complete scene or event, respectively, usually comprises a sequence of incidents, some which are very characteristic for it. An event of a certain category like e.g. a penalty or goal, a foul or the like is therefore typically defined by a very particular sequence of audio characteristics allowing an identification of the category by examining the characteristics of a succession of classified sub-units.
To identify and categorise important events in the sub-section of the source set of video signals, the key event detector examines only the audio channel of the video signals for strings of one or more classified sub-units which are similar or identical to a string of sub-units used as a definition for a certain event category, like for instance a foul. The term ‘string’ is hereby used to denote a sequence of according a some order. The comparison of the strings of classified sub-units found in the sub-section with the set of strings defining predetermined event categories yields a rating of the reliability of the respective category allocation. The rating value will be high when the string of classified sub-units is nearly identical with the string defining the event category and will decrease with an increasing mismatch between the examined and the defining string. Strings of classified sub-units which have been allocated to a certain event category are within the context of this application referred to as a segment of the video signals.
As a result of the key event detection, meta data are available which identify certain segments within the source set of video signals that correspond to certain predetermined event categories. Additionally, the reliability of the assessment of events is available by meta data representing a rating value that indicates the degree of conformity of an individual segment with the string of sub-units defining the event category assigned to the segment. The key event detector is thus capable of extracting information from a programme concerning the possible occurrence of predetermined events but is not capable of reducing the information to that what is relevant to a particular consumer.
It is therefore an object of the present invention to provide a system which is able to present to a consumer only the information contained in a certain programme which is of particular interest for the particular consumer.
This object is achieved by the invention as defined in the independent claims.
In a first embodiment of the invention, a method is proposed for defining a subset of video signals from a source set of video signals on the basis of meta data available for the source set of video signals, with the meta data assigning a generic term to a sub-section of the audio channel of the source set of video signals, a class description to one or more sub-units of the sub-section for classifying the origin of the respective sub-unit, a category allocation to a segment, which is formed by a string of one or more classified sub-units of a sub-section, and a rating value to the segment for rating the reliability of the category allocation of the segment. The method includes steps for selecting segments of a sub-section with a rating value above a defined threshold value, assigning a priority value to each category, and specifying a first subset of video signals by defining an arrangement of selected segments by an order based on the respective priority and rating values related to each segment.
In a second embodiment of the present invention an automatic programme summary apparatus is provided for defining a subset of video signals from a source set of video signals on the basis of meta data available for the source set of video signals, with the meta data assigning a generic term to a sub-section of the audio channel of the source set of video signals, a class description to one or more sub-units of the sub-section for classifying the origin of the respective sub-unit, a category allocation to a segment formed by a string of one or more classified sub-units of a sub-section, and a rating value to the segment for rating the reliability of the category allocation of the segment. The apparatus contains a selecting means for selecting individual segments from a sub-section of the source set of video signals, a meta data reading means for reading the meta data related to a selected segment, a priority assignment means for assigning a priority value to each category, a decision means for deciding if a selected segment is to be added to the subset of video signals by evaluating the rating value of the segment with respect to a threshold value, a ranking means for specifying the order in which the segments decided to be added to the subset of video signals are to be arranged, whereby the order is based on the priority value and rating value related to each respective sequence, and an arrangement means for arranging the segments decided to be added to the subset of video signals according to the order set by the ranking means.
The present invention is further implemented by a software product comprising a series of state elements which are adapted to be processed by a data processing means of an automatic programme summary apparatus according to the present invention such, that a method according to the present invention may be executed thereon.
The present invention advantageously allows to define a selection from the events detected by the key event detector based on the reliability by which a category of an event is identified and based on the priority which is assigned to the respective category of the event so that the invention yields a summary of the important events of a programme only.
Further developments of the present invention are set forth in the dependent claims.
The most important events of a programme are preferably presented first. Hereto, the arrangement of segments in the first subset of video signals is specified by order of the priority value related to each category of the segments. Within the category of a given priority, the segments are then arranged by chronological order or according to the rating value assigned to each segment.
Alternatively, the importance of the events may not primarily be defined by the priority value of the category to which the segment belongs, but by an importance value defined individually for each segment. The importance value is then formed by a weighting of the priority value related to the category of the respective segment with the rating value assigned to the segment. To achieve a presentation of events according to the assigned importance, the arrangement of the segments in the first subset of video signals is hereto specified by order of importance.
For a short overview of a programme's content, the summary is preferably limited to a certain period of time. Depending on the time limit set, only a limited selection of the important events can be presented. This is preferably achieved by selecting only those events, which can be identified with a high reliability. Therefore, the threshold value is advantageously redefined in relation to the time limit specified for the total length of the first subset of video signals.
The summary of events collected in the first subset of video signals may advantageously be restricted according to a consumer request by forming a second subset of video signals from the first subset of video signals according to a response mark obtained from the consumer on request. In a preferred embodiment of the present invention, the priority values are hereto recalculated on the basis of one or more response marks allowing to adapt future programme summaries to changing consumer preferences. The preferences of a consumer are preferably registered by means of a consumer profile which is formed on the basis of the response marks obtained from a particular consumer. Hereby, it is possible to create consumer-specific programme summaries which reflect the individual preferences of each consumer.
In the following description, the present invention is explained in more detail with respect to special embodiments and in relation to the enclosed drawings, in which:
An automatic programme summary system 12 according to the present invention is typically used within a programme suggestion engine 10 as shown in
A key event detector 11 identifies and categorises certain events of interest in a source set of video signals 13 as will be described with reference to
The present invention relies on a key event detector 11 that spots and classifies important events inside a programme forming the content of a set of video signals. As there exists no standard defining something as ‘interesting’ because the perception of a programme content is very subjective and thus differs from consumer to consumer, criteria specifying which programmes have to be considered as interesting are kept available in a consumer-specific file. For a first consumer, a respective file may e.g. assign the highest priority to international soccer games, the second highest priority to science fiction television soaps and so on, while for another consumer, the highest priority may perhaps be assigned to news magazines and the second highest priority possibly to martial art movies. A particular consumer interest is not only focused on a certain genre of programmes but very often to specific events within a programme of the genre. Such events could in the case of a soccer game e.g. be a goal, a penalty, a foul, a red or a yellow card. These events are spotted by the key event detector 11 from the video signals of a programme. The key event detection is typically based on the audio channel of the video signals as shown in
A more detailed description of the key event detection is given with reference to
To identify the predefined events, the respective strings of classes are identified in the audio stream and combined to segments as shown for the correspondingly processed audio stream 23 in
The present invention now forms a summary from the thus identified segments of key events for providing a resume for a consumer containing the interesting events. The basic steps of a method according to the invention are illustrated in the flowchart 40 of
As a result of step S01, a subset of segments is defined from the audio stream including all segments from high to low rating values. Next, in step S02 a priority value is assigned to each category which for it's part has been assigned to at least one of the segments of the subset. In the following step S03, the subset is further restricted to only those segments, which priority and rating values meet set criteria as e.g. determined in the consumer profile. The set criteria may e.g. define, that only segments of the two categories with the highest priority value, and for instance within each category only the three segments with the highest rating value are to remain in the subset. Finally, the subset of video signals is defined in step S04 by defining the arrangement of the segments selected, i.e. the segments remaining in the above defined subset, in an order according to the priority and rating values. The arrangement may be done by creating a subset of video signals containing the selected segments in the defined order, but may also be implemented in the form of meta data defining the segments selected for the arrangement and the order in which they are to be presented to a consumer.
It is to be noted that more than one threshold value can be used for selecting the segments according to a method illustrated by the flow chart 40 of
In a further embodiment of the present invention, the method for creating a subset of video signals representing a summary of a programme represented by a source set of video signals adapts the selection of segments according to a consumer input. The elementary steps of a respective method are illustrated in the flow chart 50 of
When a consumer sets a time limit for the summary in step S12, new criteria for assessing the priority and rating values of the segments according to step S03 are set in step S13 on the basis of the actual time limit set. From the given subset of video signals representing the original summary, only those segments which meet the new criteria with respect to the priority and rating values are selected in step S14 and finally presented to the consumer in step S15.
Typically, the selected segments in the subset of video signals are arranged by the order of the priority value assigned to the categories of the segments. Segments with high priority values are placed in front of segments with lower priority values. Within a given category, the segments are preferably arranged in a chronological order, as this reflects the course of events. Sometimes, a consumer is not interested to view the events in the order of their occurrence, but wants to be sure that at least the first segments presented, are correctly assigned to the category. In this case, the segments within a category are arranged by the order of the rating values assigned to each segment such, that the segments with a rating value related to a high certainty come first.
In a different approach, the arrangement of segments in the summary is specified by order of importance the events represented by the segments have for the consumer. As the meaning of importance differs from consumer to consumer, it has to be assessed for each individual consumer individually from data available in his individual profile. The interests of the consumer are best reflected in the priority values assigned to the different categories of events. Assuming, that a consumer prefers to see all events which are somehow related to his favourite category of events like e.g. foul play and goals in a soccer match, the summary will provide her or him with all segments identified to belong to her or his favourite categories; even if the category has been identified with a high uncertainty. From categories of lower priority only the events identified with a high certainty will be added to the summary. The value of importance is thus defined by weighting the priority value related to a segment with the rating value assigned to this segment. Only segments having a minimum value of importance, i.e. segments representing events that comply with the above explained importance standard, are added to the summary. Assumed, the programme for which a summary is to be arranged is a soccer match, and a consumer puts the highest priority to the category of events belonging to goals and the second highest priority to the category of events belonging to foul play, then the system would assign a higher degree of importance to goals than to foul plays. Accordingly, when defining the summary, goal scenes with a confidence score of only 0.7 would be added to the summary while for instance a foul scene with a confidence of 0.8 would be sorted out, because the consumer profile indicates that the consumer is particularly interested in goals and not so much in foul play.
The described method for defining a programme summary is implemented by an automatic programme summary apparatus 60 as e.g. shown in
The order specified by the ranking means 66 is typically given by the priority value related to the category of each segment, and within a given category the chronological order is maintained. Instead of the chronological order, the order within a category may be defined as described above by the rating value assigned to the individual segments. Alternatively, the order within a category may further be defined according to a value of importance evaluated in correspondence to what has been described above.
The value of importance is calculated by a calculating means 71 which also performs the calculations necessary for the other means of the automatic programme summary apparatus 60 like for instance the decision means 65. In
An apparatus according to the present invention, like the apparatus 60 or 60′ is typically realised by means of a data processing equipment as shown in
The present invention advantageously allows to automatically define and/or create a subset of video signals representing a summary of a programme like e.g. a TV programme or a video recording of a certain genre by using the results of a key event detector and data reflecting the preferences of a particular consumer. Changes in the preferences of a consumer can be taken into account by reassessing a consumer's preferences according to modification requests obtained by the consumer.
| Number | Date | Country | Kind |
|---|---|---|---|
| 03 026 045.9 | Nov 2003 | EP | regional |
