The embodiments discussed herein relate to a grouping apparatus, and a grouping method.
There may be a need for grouping of information provided with position information. In this case, an advertisement that is displayed on an advertising sign arranged in a certain position and information needed for operating, for example, a type of a personal computer (PC) or PC peripheral equipment in the wireless LAN environment may be examples of “information provided with position information”. “Grouping” of these pieces of information refers to putting together into one group a plurality of pieces of information provided with different pieces of position information.
For example, consider the case in which, for a series of advertisements having several types of design, a plurality of advertising signs are created for each design. A purpose of placing an advertising sign is to make an impact on a person who look at the advertising sign, that is, to provide an advertising effectiveness, so it is important to determine how to arrange the advertisement of each design. In other words, when arranging the advertisements having the same design in different positions, these advertisements may be referred to as grouped advertisements.
Further, consider that a plurality of information devices such as a PC and a printer are placed indoors, and that processing is performed that includes selecting, from among this group of information devices, a group of devices that are easily accessed from the location of a user and powering up the devices in advance. In this case, the information on the address that specifies a group of accessible devices corresponds to “grouped pieces of information”. Performing such processing permits a user of information devices to use the information devices without moving closer to the devices to power up for himself/herself.
A content data management device for managing content data that obtains, by use of the position and time that are tagged in two pieces of data from among a plurality of pieces of content data, a speed of movement between two positions in which two pieces of content data have been created and performs grouping of content data on the basis of the speed is known. For example, for image data, the speed refers to an average speed of movement when a person who captures images moves between the positions in which two images are captured. For image data, the “speed” includes the information on by which moving means the person moved to acquire a plurality of images, for example, moving by walk, moving by a car, or moving by a train. Grouping the images acquired by moving by sets of moving means whose speed is identical or similar may result in grouping more relevant pieces of content data.
A grouping apparatus is provided. The grouping apparatus includes a relationship calculation unit that calculates, at a plurality of times between a first time and a second time, the strength of correlation between each of a plurality of pieces of information and a moving body as a level of relationship for each of the plurality of pieces of information, wherein each of the plurality of pieces of information is provided to each of a plurality of objects arranged in a space and the moving body moves within the space, a similarity calculation unit that, on the basis of the level of relationship for each of the plurality of objects calculated at the plurality of times, calculates similarities for all pairs of the plurality of pieces of information at the plurality of times, integrates the similarities for all pairs of the plurality of pieces of information at the plurality of times, and calculates integrated similarity, and a classification unit that groups the plurality of pieces of information on the basis of the integrated similarity.
When grouping advertising signs, there is a possibility that the classification based on a real situation is not performed for the following reasons when only taking into account positions and directions of advertising signs. In other words, for example, signs that stand closely to one another facing in the same direction may have an advertising effectiveness significantly different from one another depending on the moving patterns of people or the spatial structure around them. Even if a classification is performed using a route distance instead of a direct distance, it is still difficult to perform a classification based on a real situation. In order to accurately conduct real-situation researches on an advertising effectiveness, behavioral patterns of people and a spatial structure also need to be considered when classifying signs.
Consider that a plurality of information devices such as a PC and a printer are placed indoors, and that processing is performed that includes selecting, from among this group of information devices, a group of devices that are easily accessed by a person in a certain location and powering up the devices in advance. In this case, the devices adjacent to one another across the wall may be powered up when a group of devices simply grouped by use of a direct distance is turned ON. If a route distance is used, grouping may be performed according to different criteria than the accessibility of people, depending on the relationship between the positions of the devices.
A grouping apparatus and a grouping method according to an embodiment will now be described with reference to the drawings.
<General Description>
First, referring to
In this case, “information provided with position information” may be information that is arranged in a position in a three-dimensional space that is a living space of a person. A personal computer (PC) or PC peripheral equipment in the wireless LAN environment of advertisements, sounds, and images are examples of such information. “Grouping” of such information indicates a distribution of a plurality of pieces of information provided with different pieces of position information to a plurality of groups.
For example, consider that a plurality of information devices such as a PC and a printer are placed indoors, and that processing is performed that includes selecting, from among this group of information devices, a group of devices that are easily accessed by a person in a certain location and powering up the devices in advance. Each information device has information provided with position information. For example, each information device has information such as an IP address and a MAC address that identifies itself, and configures “information provided with position information” by combining its information with the information on the position in which the information device is placed.
In this case, “group of easily accessible devices” corresponds to grouped information devices. Performing such processing permits a user of information devices to use the information devices without moving closer to the devices to power up for himself/herself.
Further, “information” may be, for example, content data such as image data, audio data, and text data that are stored in an information recording medium as an electronic file.
First, referring to
An office illustrated in
In
Group A: information device 104 (room R1), information device 106 (room R3)
Group B: information device 102 (room R1), information device 108 (room R2)
Group C: information device 112 (passage R1), information device 114 (passage P1), information device 110 (room R2), and information device 116 (room R3)
The information devices 102 and 108 that belong to Group 2 above are arranged in different rooms, and the direct distance between them goes through a wall. Even if a communication between information devices is wireless, there may be a great attenuation of signal intensity depending on the structure of the wall when going through a wall, so it is difficult to consider classifying the information devices 102 and 108 into the same group as “natural” grouping. Thus, there is a possibility that grouping by use of a direct distance between information devices is not “natural”.
In
When grouping on the basis of a route distance, the following two groups are formed:
Group D: information device 104 (room R1), information device 106 (room R3)
Group E: information device 112 (passage P1), information device 114 (passage R1), information device 110 (room R2), information device 116 (room R3)
In this case, the route distance between the information device 102 (room R1) and the information device 108 (room R2) of Group B passes through the passage P1, and is not small enough for the information devices 102 and 108 to be grouped. The information devices 104 and 106 are arranged in the rooms R1 and R3, respectively, but the route distance between them is not so different from the direct distance between them. Thus, the information devices 104 and 106 are also grouped when it is determined whether they are to be grouped depending on the route distance.
As described above, when grouping on the basis of the route distance, the information devices arranged in the different rooms are also put together into the same group. Thus, there is a possibility that grouping by use of a route distance between information devices is not “natural”.
In
Group F (section A): information device 106, information device 116, and information device 118
Group G (section B): information device 110, and information device 114
When performing grouping for each of the predetermined sections in this way, an additional work will be needed for classification and the method for determining a section may have arbitrariness.
In
Group H (room R1): information device 102, and information device 104
Group I (room R2): information device 108, and information device 110
Group J (passage P1): information device 112, and information device 114
Group K (room R3): information device 106, information device 116, and information device 118
In this case, the information devices to be grouped are in the same room, and there is not any wall between the information devices. Further, the route distances from one another are short because the information devices to be grouped are arranged in the same room.
Classifying according to a distance between information devices is not sufficient in order to perform natural grouping as illustrated in
In
It is assumed that, for a series of advertisements having several types of design, there are a plurality of advertising signs for each design. “Information provided with position information” may be a design of advertisement that is provided with the information about the position of the advertising sign.
For the plurality of advertising signs, the advertising signs that are provided with the advertisements having a similar advertising effectiveness to one another may be placed in one group. The advertising effectiveness may be an impact that is made on a person who looks at the advertising sign. Advertisements having the same design correspond to “grouped pieces of information”. A purpose of placing an advertising sign is to make an impact on a person who looks at the advertising sign, so it is important to determine how to arrange advertising signs having the same design.
In
A dotted line indicates a trajectory of a moving body 200 that moves in the space in which the advertising signs are arranged. The moving body 200 can be a person, a vehicle, or anything that moves recognizing the advertisements provided to the advertising signs. When the moving body is a vehicle, the advertisements may be traffic signs.
In
A moving body is able to recognize an advertisement provided to an advertising sign that is included within its field of view. For example, the moving body is able to recognize the advertisements provided to the advertising signs 304 and 306 at the time t=T1. Thus, when performing natural grouping of information, the information, that is, the advertisement provided to the advertising sign 304, and the advertisement provided to the advertising sign 306 are grouped.
Further, the moving body 200 is in the position x=P2 at the time t=T2. When the moving body 200 that was in the position x=P1 at the time t=T1 arrives at the position x=P2, the advertising signs that are included within the field of view of the moving body and whose distance from the moving body 200 is less than a predetermined value are the advertising signs 314, 316, and 318. Further, the advertisements provided to the advertising signs 314, 316, and 318 have distances not greater than the predetermined value (second threshold) from one another. Thus, when performing natural grouping of information, the advertisements provided to these advertising signs 314, 316, and 318 are grouped.
In
As described above, it may be important to consider the movement of an information recipient when performing natural grouping of information. In other words, it is preferable to group pieces of information whose level of accessibility at the same time is similar as much as possible. Further, it is preferable to group pieces of information that are correlated with the information user as long as possible.
The moving body 400 is in the position x=P4 at the time t=T4. At that time, the advertising signs that are included within the field of view of the moving body 400 and whose distance from the moving body 400 is less than the predetermined value (first threshold) are the advertisements provided to the advertising signs 506 and 508. Further, a distance between the advertisements provided to the advertising signs 506 and 508 is not greater than the predetermined value (second threshold) from each other. Thus, when performing natural grouping of information, the information provided to these advertising signs 506 and 508 are grouped.
When the moving body 400 that was in the position x=P4 at the time t=T4 has moved to the position x=P5 at the time T5. At that time, the advertisements that are included within the field of view of the moving body 400 and whose distance from the moving body 400 is less than the predetermined value (first threshold) are the advertisements provided to the advertising signs 516 and 518 arranged in the passage P2. Further, the distance between the advertisement provided to the advertising sign 516 and the advertisement provided to the advertising sign 518 is less than the predetermined value (second threshold). Thus, when performing natural grouping of information, the information provided to these advertising signs 516 and 518 are grouped.
When the moving body 400 that was in the position x=P5 at the time t=T5 has moved to the position x=P6 at the time T6. At that time, the advertising signs that are included within the field of view of the moving body 400 and whose distance from the moving body 400 is less than the predetermined value (first threshold) are the advertising signs 526, 528 and 530. The advertisement provided to the advertising signs 534 is included in the field of view but whose distance from the moving body 400 is greater than the predetermined value (first threshold). Further, the advertisements provided to the advertising signs 526, 528, and 530 have a distance less than the predetermined value (second threshold) from one another. Thus, when performing natural grouping of information, the information provided to these advertising signs 526, 528, and 530 are grouped.
As described above, when grouping advertising signs that have a similar advertising effectiveness to one another in order to conduct researches on an advertising effectiveness of a plurality of advertising signs, there is a possibility that the classification based on a real situation is not performed when only taking into account positions and directions of advertising signs. For example, signs that stand closely to one another facing in the same direction may have an advertising effectiveness significantly different from one another depending on the moving patterns of people or the spatial structure around them. Even if a classification is performed using a route distance instead of a direct distance, it is still difficult to perform a classification based on a real situation. In order to accurately conduct real-situation researches on an advertising effectiveness, the behavioral pattern of a person who is a target of the advertisement and the spatial structure also need to be considered when classifying signs.
“Natural grouping of information” is related to the accessibility of an information user to information at the same time. In this case, the accessibility to information is related to an amount of additional movements to be performed by a user or a recipient of the information in order to access the information. The movement of people or vehicles has inertia, and their heading direction is often the same as the center of their field of view. On the other hand, a large additional movement may be needed when suddenly changing directions or when looking at a direction that is different from the heading direction. Further, the access is easier if a distance between a moving body and an object provided with information is shorter. The reason that the accessibility to information at the same time is important is that an information user (recipient) may be in different locations at different times, and it is not easy for the user to associate the pieces of information that were recognized in the different locations at the different times.
In this case, when the information is an advertisement and the user is a person, for example, a “user accesses information” may refer to the state in which the advertisement is included in the field of view of that person. Further, in the wireless LAN environment when the information is a printer and the user is a PC user, it may refer to the state in which the printer operates according to the instructions from the PC.
Further, when performing “natural grouping of information”, the accessibility of a user (moving body) to a plurality of pieces of information at a certain time is considered. The equality of the accessibility of a user to a plurality of pieces of information at a certain time may be referred to as “simultaneity of the accessibility to information”. For example, the accessibility can be determined only using the distance between an information user and information provided to an object, that is, the distance between the user and the object. For example, consider that an object A and an object B are away from each other, and that a distance between a moving body and information provided to the object A and a distance between the moving body and information provided to the object B are the same at different times. In this case, if there is no need to consider the times, it is possible to equally access the information provided to the object A and the information provided to the object B. However, in reality, they happen at the different times, and the information provided to the object A and the information provided to the object B are not grouped when performing “natural grouping of information” because their “simultaneity of the accessibility to information” is low.
Further, if a moving user stays longer in an area in which certain information is accessible, he/she can access the information more easily. This may be referred to as “continuity of the accessibility to information”. Here, consider two pieces of information A and B. It is assumed that, for a user, the amount of time in which the information A can be accessed and the amount of time in which the information B can be accessed are the same. However, if their time periods do not have any overlapping portion, the information A and the information B are not to be grouped for the user.
As described above, when performing “natural grouping of information provided with position information”, the pieces of information whose level of “simultaneity of the accessibility” and level of “continuity of accessibility” are both high are grouped.
In other words, “natural grouping of information provided with position information” may be defined as follows:
“Putting together into one group a plurality of pieces of information to which the accessibility of an information user at the same time is high and that are correlated with the information user as long as possible” . . . . (*)
Further, this definition can be paraphrased as follows:
“Putting together into one group a plurality of pieces of information whose simultaneity and continuity of accessibility of an information user are both high” . . . . (**)
A grouping apparatus that will now be described performs following processing so as to perform natural grouping that is not determined not only using the position of the information, and the distance between a moving body that receives the information and the information. The strength relationship between moving bodies 200 and 400 that move in a space in which pieces of information provided with position information exist and each of the pieces of information is obtained at a certain time. The strength of relationship between the moving bodies 200 and 400 and each of the pieces of information is designed to change according to the position, direction, and speed of the moving bodies 200 and 400. Next, with respect to the relationships between the moving bodies 200 and 400 and any two of the pieces of information, similarity that indicates the strength of correlation between the two relationships is obtained, and an integrated similarity set is created by integrating similarities every time the state of the moving bodies 200 and 400 changes and by normalizing it. Then, an information classification is performed using the integrated similarity set. Using the grouping apparatus that is configured in this way permits an improvement of the efficiency of an information transmission to an information user because a plurality of pieces of information whose level of accessibility of the information user at the same time is similar and that are correlated with the information user as long as possible are put together into one group.
<Grouping Apparatus>
A grouping apparatus 700 according to an embodiment will now be described with reference to
A sensor 600 is connected to a grouping apparatus 700. The sensor 600 and the grouping apparatus 700 are combined to configure a grouping system.
The sensor 600 is attached to a moving body 650. The moving body 650 may be, for example, the moving body 200 in
The grouping apparatus 700 includes a position measurement unit 702, a relationship calculation unit 704, an object data storing unit 706, a similarity calculation unit 708, and a classification unit 710.
The position measurement unit 702 receives data about a position of the moving body 650 that has been measured by the sensor 600. Further, the position measuring unit 702 receives data about a speed of the moving body when the sensor 600 measures the speed of the moving body. Then, the position measurement unit 702 determines a position and a direction of the moving body.
Further, the position measurement unit 702 measures time. The time measured by the position measurement unit 702 can also be shared by the relationship calculation unit 704, the object data storing unit 706, the similarity calculation unit 708, and the classification unit 710. However, the measurement of time may be performed by a unit other than the position measurement unit 702. For example, time may be measured by a clock that is not illustrated. In this case, the time measured by the clock can also be shared by the position measurement unit 702, the relationship calculation unit 704, the object data storing unit 706, the similarity calculation unit 708, and the classification unit 710.
The object data storing unit 706 stores therein data about an object provided with “positioned information”. The data may include a position of the object arranged in a space in which the moving body moves, and an orientation of the information. Further, when there is an obstacle such as a wall in the space, the object data storing unit 706 may store therein the information on a position and a material of the obstacle.
Referring to the data stored in the object data storing unit 706, the relationship calculation unit 704 calculates a relationship between the information provided to each object and the moving body 650.
For example, when the position of a certain object is Xk, the position of the moving body 650 is Y, and the monotone decreasing function is F(x), the relationship rk between the information provided to the object positioned in Xk and the moving body 650 is defined using
r
k
=F(|Xk−Y|) (1).
r
k=1−α·dk(dk<1/α)
r
k=0(dk≧1/α) (3).
A nearest-neighbor model will now be described.
In
The Laplacian potential is most generally used to represent potential. The Laplacian potential can be represented as a solution of the Laplace equation such as
Δφ(x)=δ(x−Y) (4).
The δ function is a value when a parameter is zero, but otherwise it is zero. The solution of the Laplace equation has the following features and advantages. The solution of the Laplace equation is continuous and smooth, and has no extreme values except for a maximum point. Thus, the solution of the Laplace equation is suitable for naturally representing the shape that extends with the moving body in its center avoiding the wall, by generating a distribution in which the maximum value is the existing position of the moving body and the minimum value is the wall surface. Further, the solution of the Laplace equation is a function of location, so there is an amount (elevation of potential) at any point on a map. This is the amount that decreases with distance from the position of the moving body (center of potential), so that is suitable to be used as an indicator of distance in which a route of the moving body is considered. The Laplacian potential can be used as a method for representing a near-field area because of these advantages. It is possible to take into account the existence of an obstacle such as a wall by setting the requirement that the solution of the Laplace equation satisfies the boundary condition φ=0 on the surface of an obstacle.
In this case, using a solution φ(x) of the Laplace equation, the relationship rk between the information provided to the object positioned in Xk and the moving body 650 may be defined as
r
k=φ(Xk) (5).
In other words, when an equal level contour of φ(x) of the Laplace equation is described, it is related to the distance between the object positioned in Xk along the line perpendicular to the equal level contour and the moving body 650. In this case, the solution of the Laplace equation satisfies the boundary condition φ=0 on the surface of the wall, so the line perpendicular to the equal level contour does not pass through the obstacle.
When numerically solving a partial differential equation such as Formula 4 above, the relationship calculation unit 704 performs a calculation by dividing space into grid, but the Laplace equation is a partial differential equation of elliptic type and it is not possible to explicitly obtain values of cells of the grid, so a large simultaneous equation whose variable is the number of cells of the grid obtained by the division has to be solved. It takes much time to calculate this. The number of cells of the grid (resolution of space) tends to increase in order to represent the potential distributed in a space including an obstacle such as a wall. Further, considering that a simultaneous calculation may be performed with respect to a plurality of moving bodies, it is not possible to use a general grid solution because it needs too much time.
Thus, the relationship calculation unit 704 performs a potential generating calculation using a charge simulation method in order to deal with a problem of the calculation time. Using the charge simulation method permits reducing of the time needed for a potential generation.
The charge simulation method is an approximate solution method for a partial differential equation. The charge simulation method is a simple algorithm, so it permits shortening of the calculation time needed for obtaining an approximate solution and permits obtaining of a relatively high accuracy of approximation with respect to a smooth boundary. However, when the boundary shape is complicated, it is difficult to use the charge simulation method because the number of errors increases. The algorithm of the charge simulation method will now be described.
As described above, in the charge simulation method, a sample point and a charge point are determined. In the charge simulation method, the following processes are performed:
(Process 1) Set an analysis area for which potential is obtained as an area Ω, and select sample points Xi (i=1, 2, . . . , n: n is a natural number) inside the area Ω.
(Process 2) Designate a value bi of the potential that corresponds to the position of each of the selected sample points Xi.
(Process 3) Select charge points Yi (i=1, 2, . . . , n: n is a natural number) outside the area Ω.
(Process 4) Assume a solution u(X) such as Formula 6 below, and determine a coefficient Qi using Formula 2 so that the solution u(Xi) satisfies the designated condition bi.
In this case, |X−Yi| indicates the distance between the sample point X and the charge point Yi.
Eventually, this simultaneous equation also needs to be solved in the charge simulation method. However, the time needed for calculation is much shorter because the number of unknowns is fewer compared to the conventional grid solution.
In the charge simulation method, selecting several charge and sample points along a boundary of a space to be analyzed permits calculating of the potential that is smoothly distributed with respect to the boundary and that satisfies the designated condition on the boundary. Using such characteristics, the relationship calculation unit 704 of the grouping apparatus 700 is able to generate a potential distribution that extends along the passable area within the space by arranging the sample points and the corresponding charge points along the surface of the obstacle.
In
The case in which the relationship rk is determined only using the position of the object provided with information and the position of the moving body 650 has been described above, but the relationship calculation unit 704 may consider the orientation of the information, or the speed or the orientation of the moving body 650 to calculate the relationship rk.
For example, it is assumed that the position of an object provided with information is Xk and the orientation of the information is Vk. Further, it is assumed that the position of the moving body 650 is Y and its orientation is VY. The orientations V1 and VY may be unit vectors of magnitude one.
In this case, the relationship calculation unit 704 may define the relationship rk between the information provided to the object positioned in Xk and the moving body 650 using
The value of this formula is larger if the distance between the information provided to the object positioned in Xk and the moving body is smaller and when the orientations of the information and the moving body face each other. The value of this formula is also large when the moving body and the information face in opposite directions. In this case, it is possible to calculate a cross product of the orientations Vk and VY, and to determine, from a sign of a result of the calculation, the case in which they face in the directions exactly opposite to each other, and it may be set to rk=0.
Further, the relationship calculation unit 704 may define the relationship rk between the information provided to the object positioned in Xk and the moving body 650 using
r
k=(1−Vk·VY)×φ(Xk) (9)
by use of the potential φ(x) of Formula 5. In this case, the relationship rk is larger if the distance along the line perpendicular to the potential is shorter, and is large when the orientation Vk of the information provided to the object and the orientation VY of the moving body face each other.
In
After the relationship calculation unit 704 calculates the relationship rk between the information provided to the object positioned in Xk and the moving body 650, the similarity calculation unit 708 calculates similarities C(i,j) with respect to all pairs of pieces of information. A collection {C(i,j)} of the similarities C(i,j) with respect to all pairs of pieces of information may be referred to as a similarity set. For example, the similarity calculation unit 70 calculates a similarity C(i,j) between the information provided to the object positioned in Xi and the information provided to the object positioned in Xj. If the value of the similarity C(i,j) is larger, the information provided to the object positioned in Xi and the information provided to the object positioned in Xj are “more similar”, and they are more likely to be grouped. If there are n objects, the similarity C(i,j) has (½)n(n−1) independent components. A set of (½)n(n−1) similarities may be referred to as a similarity set.
For example, similarity C(i,j) between information provided to the object positioned in Xi and information provided to the object positioned in Xj preferably has characteristics such that, using a relationship ri between the information provided to an object positioned in Xi and a moving body positioned in Y, and a relationship rj between the information provided to an object positioned in Xj and the moving body positioned in Y,
(C1) symmetric with respect to interchanging between the indexes i and j, that is, C(i,j)=C(j,i),
(C2) similarity C(i,j) is zero when either of the relationships is zero, and
(C3) the value is larger when the values of ri and rj are larger and the difference between them is smaller.
For example, similarity C(i,j) at a certain time t may be defined as
The value of this formula is larger if the values of the relationship ri and the relationship rj are larger and the difference between them is smaller. The similarity C(i,j) is zero when either of the relationships is zero. The similarity C(i,j) has the symmetric characteristics with respect to interchanging between the indexes i and j, so C(i,j)=C(j,i). A similarity set can be expressed in a matrix form, as
A matrix C represents similarity of a relationship with information provided to an object when a moving body is in a certain position Y.
The similarity C(i,j) between the information provided to the object in the position Xi and the information provided to the object in the position Xj when the moving body is in the position Y has been defined above using Formula 5, but it may be defined using another formula that satisfies (C1) to (C3).
This similarity C(i,j) indicates to what extent the information provided to the object in the position Xi and the information provided to the object in the position Xj can equally be recognized from the moving body when the moving body is in the position Y at a certain time t.
Further, the similarity calculation unit 708 calculates a similarity set that is represented using a matrix C every time the moving body moves and integrates them, so as to calculate integrated similarity.
For example, if a similarity set at a time t is C(t), integrated similarity C* is
C*=∫dtC(t) (12).
The integrated similarity C* is a matrix that includes C*(i,j) as an element. In this case, integrals with respect to the matrix are obtained by performing an integration for each matrix element and are expressed in a matrix form, that is, they are defined as
Formulas 8 and 9 describe as if time is a continuous variable, but in fact, the sensor 600 acquires the information such as the position and the orientation of the moving body 650 discretely with respect to time. Thus, in Formulas 8 and 9 above, a discrete system may be approximately represented by use of a continuous variable t, as in Formula 16 below.
The integral range in Formulas 8 and 9 above is from a time to another time, but it may be, for example, a time period during which a moving body arrives at a position from another position.
Further, the matrix C* may be normalized by dividing each element by a largest element. The moving pattern of the moving body is reflected in the matrix C* of similarities obtained by continually integrating during a certain time period. The matrix C* provides a set of integrated similarities.
The example of one moving body has been described above. However, the above description may also be readily expanded to the case in which there are a plurality of moving bodies.
The classification unit 710 groups pieces of information on the basis of the similarity calculated by the similarity calculation unit 708. The classification unit 710 creates a graph of relationships linked to one another by use of links whose strength is represented by an element value in the similarity matrix C*, and performs grouping by severing the links of the elements whose linkage strength is not greater than a threshold.
The table of
The classification unit 710 severs, for example, the links in the graph of
When the number of pieces of information is large, the size of the similarity matrix C* of Formula 9 becomes large accordingly.
In
The table of
In the theory of hierarchical clustering, for example, the following processing is performed. When the data formed with N targets x1 to xN is provided, a state in which there are N clusters only including one target is created as an initial state. A distance d(C1,C2) between clusters is calculated from a distance d(xi,xj) between the target xi and the target xj (dissimilarity), and the two clusters having a closest distance from each other are sequentially merged. A result of such processing can be represented by a dendrogram. The dendrogram is a binary tree in which each non-terminating node represents each target and a cluster obtained by merging is represented by use of non-terminating nodes. There are several methods depending on the functions that represent a distance d(C1,C2) between clusters. For example, in the nearest neighbor method, the distance d(C1,C2) between clusters is defined as a minimum distance from among the distances between the targets included in the cluster C1 and the targets included in the cluster C2. In the furthest neighbor method, the distance d(C1,C2) between clusters is defined as a maximum distance from among the distances between the targets included in the cluster C1 and the targets included in the cluster C2. Further, the group average method that defines, as the distance d(C1,C2) between clusters, an average of the distances between the targets included in the cluster C1 and the targets included in the cluster C2, and the Ward's method that minimizes the sum of squares of distances from the targets to the centroids of the clusters including the respective targets are known. Any of the methods above can be used for hierarchical grouping.
For example, with respect to the pieces of information A, B, and C in
As illustrated in
Further, for example, as illustrated in
As described above, at the time t1, the levels of relationship between the moving body and each of the pieces of information provided to the objects 1102, 1104, 1106, and 1108 that are arranged in the same room R8 is high. On the other hand, at the time t2 when the moving body has moved to the room 9, the levels of relationship between the moving body and each of the pieces of information provided to the objects 1102, 1104, 1106, and 1108 that are arranged in the room R8 is low. Accordingly, when a moving body and information exist in positions close to each other at a certain time, the level of their relationship r is high.
As described above, with respect to similarity C, similarity C(i,j) between information positioned in Xi and information positioned in Xj from among a plurality of pieces of information has characteristics such that, using a relationship ri between the information positioned in Xi and a moving body positioned in Y and the relationship rj between the information positioned in Xj and the moving body positioned in Y,
(C1) C(i,j)=C(j,i), that is, symmetric with respect to interchanging between the indexes i and j,
(C2) C(i,j)=0 when either of the relationship ri or the relationship rj is zero, and
(C3) the value is larger when the values of the relationships ri and rj are larger and the difference between the relationships ri and rj is smaller.
For example, with respect to the objects 1102 and 1104 in
Consider integrated similarity C* obtained by time-integrating the relationship r. An element in the integrated similarity C* is larger if a time period during which an object provided with information is included in the field of view of a moving body is longer.
According to such a grouping apparatus 700 as described above, “natural grouping of information” based on a movement route of an information user and the accessibility to the information at a certain time can be performed.
For example, when the object is an advertising sign and the information is an advertisement, the classification is based not only on the level of advertising effectiveness but also on the simultaneity. In other words, the advertising signs that increase the advertising effectiveness simultaneously are to be put together considering the behavioral patterns of people who look at the advertisement. According to this classification scheme, if there are signs that are physically close to each other and stand facing in the same direction and if there is an increase in advertising effectiveness for each of the signs, the two signs are not to be grouped unless there is a moment when there is an increase in advertising effectiveness for the two advertising signs at the same time. It will be appreciated that, in other words, signs that are physically similar may belong to different groups by interpreting as inclusive of behavioral patterns of information users and the spatial structure from the perspective of advertising effectiveness.
In
As illustrated in
The computer 800 includes an MPU 802, a ROM 804, a RAM 806, a hard disk device 808, an input device 810, a display 812, an interface device 814, and a recording medium driving device 816. These components are connected to one another via a bus line 820, and are able to transmit and receive various pieces of data with one another under the control of the MPU 802.
The MPU (micro processing unit) 802 is an arithmetic processing unit that controls the operation of the entire computer 800 and serves as a control processing unit of the computer 800.
The ROM (read only memory) 804 is a read only semiconductor memory that has recorded thereon a predetermined basic control program in advance. The MPU 802 reads and executes the basic control program when the computer 800 starts, and thereby it is possible to control the operation of each component of the computer 800.
The RAM (random access memory) 806 is a semiconductor memory that is recordable and readable at any time and which the MPU 802 uses as a working storage area as needed when the MPU 802 executes various control programs.
The hard disk device 808 is a storage that stores therein various programs and data executed by the MPU 802. The MPU 802 is able to perform a variety of various control processes that will be described below, by reading and executing the predetermined control programs stored in the hard disk device 808.
The input device 810 is, for example, a mouse device or a keyboard device, and when it is operated by a user of the grouping apparatus 700, it acquires an input of various pieces of information associated with the operational contexts and transmits the acquired input information to the MPU 802.
The display 812 is, for example, a liquid-crystal display, and displays various texts and images according to the display data transmitted from the MPU 802.
The interface device 814 manages a transmission and a reception of various pieces of information between various devices that are connected to the computer 800.
The recording medium driving device 816 is a device that reads various programs and data recorded on a portable recording medium 818. The MPU 802 is also able to perform the various control processes that will be described below, by reading and executing the predetermined control program stored in the portable recording medium 818 via the recording medium driving device 816. A flash memory, a CD-ROM (compact disc read only memory), and a DVD-ROM (digital versatile disc read only memory) that are equipped with a USB (universal serial bus) standard connector are examples of the portable recording medium 818.
In order to configure the grouping apparatus 700 by use of such a computer 800, for example, a control program is created that causes the MPU 802 to perform the processing in each of the processing devices described above. The created control program is stored in the hard disk device 808 or in the portable recording medium 818 in advance. Then, a predetermined instruction is given to the MPU 802 so that the control program is read and executed. This permits the MPU 802 to provide the functions included in the grouping apparatus 700.
As described above, the grouping apparatus 700 permits an improvement of the efficiency of an information transmission to an information user because a plurality of pieces of information whose level of accessibility of the information user at the same time is similar and that are correlated with the information user as long as possible are put together into one group.
<Grouping Processing>
When the grouping apparatus 700 is the general-purpose computer 800 illustrated in
When the processing starts, the position measurement unit 702 of the grouping apparatus 700 resets the time to t=0 in S102. Then, when the process in this step ends, the process moves on to S104.
In S104, the position measurement unit 702 of the grouping apparatus 700 advances time by Δt. Then, when the process in this step ends, the process moves on to S106.
In S106, the position measurement unit 702 of the grouping apparatus 700 measures a position of a moving body. Here, it may measure a direction, that is, a heading direction of the moving body. Then, the process moves on to S108.
In S108, the relationship calculation unit 704 of the grouping apparatus 700 calculates a relationship between each object and the moving body referring to the data stored in the object data storing unit 706. Then, when the process in this step ends, the process moves on to S108.
For example, when it is assumed that a position of a certain object is Xk, a position of a moving body 650 is Y, and a monotone decreasing function is F(x), a relationship rk between the object positioned in Xk and the moving body 650 is calculated using
r
k
=F(|Xk−Y|) (14).
In particular, Formula 13 described above may be any of Formula 2, Formula 3, or Formula 5. Then, when the process in this step ends, the process moves on to S110.
In S110, the similarity calculation unit 708 of the grouping apparatus 700 calculates similarity C(i,j) with respect to all pairs of objects.
Using the relationship ri between an object positioned in Xi and an moving body positioned in Y, and the relationship rj between an object positioned in Xj and the moving body positioned in Y, similarity C(i,j) between information provided to the object positioned in Xi and information provided to the object positioned in Xj is preferably a function that satisfies (C1) to (C3) described above. As an example, the similarity C(i,j) may be
as described above.
A collection {C(i,j) (t)} of the similarities C(i,j) (t) at a time t with respect to all pairs of objects may be referred to as a similarity set. Then, when the process in this step ends, the process moves on to S112.
In S112, the similarity calculation unit 708 of the grouping apparatus 700 calculates integrated similarity C* at a time t by adding the similarities C(i,j) calculated in S110 to the existing similarity set, that is,
C*(t)=C*(t−Δt)+C(t)Δt (16)
is calculated. In this case, C(t) is a matrix, so Formula 15 described above is a matrix equation. Then, when the process in this step ends, the process moves on to S114.
In S114, the similarity calculation unit 708 of the grouping apparatus 700 normalizes the integrated similarity C*. For the way of normalizing, for example, each element may be divided by a maximum matrix element of the integrated similarity C*. When the process in this step ends, the process moves on to S116.
In S116, the classification unit 710 of the grouping apparatus 700 performs grouping information on the basis of the similarities calculated in S110 to S114. For example, it creates a graph of relationships linked to one another by use of links whose strength is represented by an element value in the similarity matrix C*, and performs grouping by severing the links of the elements whose linkage strength is not greater than a threshold.
In S118, the classification unit 710 of the grouping apparatus 700 determines whether the termination condition is satisfied. The termination condition may be whether a specific moving body has arrived at a certain position. The termination condition may also be whether a predetermined time period has elapsed. Further, the process in this step may be performed by a component other than the classification unit 710, for example, by the position measurement unit 702. If a determination result in this step is “Yes”, that is, if it satisfies the termination condition, the processing terminates. If the determination result in this step is “No”, that is, if it does not satisfy the termination condition, the process returns to S104.
Performing such processing permits natural grouping of information provided with position information such as (*) and (**) described above to be performed. It is possible to efficiently report information to an information user because grouping of information is performed on the basis of the accessibility of the information user to pieces of information at the same time.
According to an aspect of the embodiments, a plurality of pieces of information can be grouped so as to improve the efficiency of an information transmission to a moving body.
This application is a continuation application of International Application PCT/JP2013/063378 filed on May 14, 2013, and designated the U.S., the entire contents of which are incorporated herein by reference.
Number | Date | Country | |
---|---|---|---|
Parent | PCT/JP2013/063378 | May 2013 | US |
Child | 14924093 | US |