JPH0723012A - Audience rating survey system - Google Patents

Abstract

PURPOSE:To firmly establish the system for generating a precese audience rating without depending on reports of TV viewers. CONSTITUTION:In this system, a television camera 3 for photographing an audience space of a television 1 is installed, and this system is provided with a segmenting part 5 for extracting discriminating information of a TV viewer at every image frame unit from a video signal of the camera 3, its storage part (memories 6, 7), a face recognizing part 8 for extracting a face image from a signal of the memory 6 and imparting an identification number, an overall deciding part 9 for deciding a state in the past and determining personal discriminating information, and a clock for obtaining the information for showing the time at that time therein, and provided with a transfer means for sending channel information of the television 1, and perfonal time zone information by the abovementioned personal discriminating information to an information processor. A person who views the television 1 is specified in advance and stored in advance as the personal discriminating information. Subsequently, from an image frame of a person who is looking at actually, the face image is extracted and by the overall deciding part 9, a specific person can be identified.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audience rating survey system for detecting and tabulating trends of viewers who receive television broadcasting.

[0002]

2. Description of the Related Art Conventionally, a television receiver (hereinafter referred to as a television) has been used for conducting a television broadcast audience rating survey.
It required channel information such as the operating occupation time and which channel was selected. The channel information is obtained by interlocking with the channel selection function or by detecting the tuning frequency generated inside the television. The occupation time information and the channel information are output at constant time intervals, information processing is performed, and the audience rating is investigated.

With such a survey device, for example, it may not be possible to know who is watching the family at home, and it may be the infant who was watching the program for women, or in extreme cases. There were some problems, such as the TV being on and the room unattended.

There is also a method of submitting a diary to grasp the viewing situation for each individual, but there is a problem that the entered data is not always accurate. As a device that mechanically measures the viewing situation for each individual, there is a device called a people meter, and each viewer has a switch, and each person opens and closes the switch at the start and end of program viewing. Although there are devices that perform measurement by using, there are many cases where the switch is forgotten and accurate data cannot always be collected.

Therefore, in order not to rely on the viewer's declaration, the TV is equipped with a TV camera, the interior of the TV is photographed by the TV camera, and it is judged from the captured image whether or not the TV is viewed. It only needs to be able to recognize who a person is. In order to identify each viewer, it is possible to target the face watching TV.

[0006] Regarding face determination, there are many studies on recognition from a still photograph of a front face. For example, the magazine "Measurement and Control"
Commentary article "Face identification technology" published by Satoshi Minami, 19
Published in August 1986). In addition, in recent research, "Electronic Information and Communication Engineers Transactions", J73-D-II, Vol. 9, "Automatic identification of human face using profile profile obtained from three-dimensional data" (Toru Abe) Other work, 1990
(September 2013), the shape of the face was measured three-dimensionally,
The face data obtained by this measurement is three-dimensionally rotated in a computer, a profile image is obtained, and a method for recognizing an individual by this profile image is proposed, and a high recognition rate is obtained.

[0007]

However, when face recognition is performed, in the above-mentioned front still image, the viewer must face directly in front of the television, and a moving image based on the fact that the viewer moves. There were not many attempts from the statue,
This is not suitable for practical use, since the viewer may be facing sideways or the face may be hidden in a newspaper. In addition, although three-dimensional measurement can be made highly accurate and discrimination ability can be improved,
As a preparatory step, it is necessary to illuminate with slit light with the surroundings darkened, and it takes too much time for processing, which is difficult to use for a television viewing situation survey.

The present invention recognizes a viewer with a television camera, and reliably identifies a plurality of viewers in front of the television even if the viewer is moving around. It is an object of the present invention to provide an audience rating survey system capable of demanding a high audience rating.

[0009]

In order to achieve the above-mentioned object, the present invention provides a television receiver, a viewer image input means directed to a viewing space of the television receiver, and the viewer image input means. Means for obtaining the identification information of the viewer cut out from the image frame extracted by, means for storing the identification information of the specific viewer in advance, and identification information of the viewer newly extracted by the viewer image input means Determination means for comparing and identifying the identification information of the specific viewer stored by the storage means for detecting the time zone of the input signal of the identification information of the viewer newly added to the determination means. A clock means and a transmission means for transmitting the personal time zone information obtained by the determination means and the channel information of the television receiver to the information processing apparatus are provided. The image frame corresponds to a signal element that constitutes one screen.

Further, when a plurality of extracted identification information are confirmed at consecutive positions between a plurality of consecutively extracted image frames, the same identification number is given to the consecutive plurality of identification information and stored. Storage means, determination means for identifying a new person who has been given an identification number and a specific viewer, and replacing the identification number or stored information of the person who has been given the identification number as identification information for the specified viewer. To provide. Alternatively, in the determining means, the identification information of the unidentified identification number is sent to the information processing device, and when it is determined that the person to whom the unidentified identification number is given is the same as the identified viewer, the transmitted personal time It is characterized in that a means for transmitting a signal for associating the band information with the subsequently identified personal time band information is provided.

Further, the above system is characterized in that it is provided with means for calibrating the time of the clock means in response to a signal from the information processing side.

In addition, the number of persons existing in the viewing range is counted from the image frame, or the viewing mode is identified, and the number-of-people information or the information indicating the viewing mode is transmitted together with the personal time zone information. Is characterized by.

Further, in order to obtain the above-mentioned identification information to be stored, the identification information of a specific person is given an identification number to be learned and stored, and the identification information to which the identification number is attached is specified for viewing. It is characterized in that it is replaced with the identification information of each person, and that each person's position confirmation device is provided, and the image of each person in the screen frame is identified from the output of the position confirmation device during learning. .

Furthermore, in these systems, a plurality of such systems are provided, a plurality of systems according to claim 1 are installed, and personal time zone information and channel information for a plurality of television receivers are comprehensively determined, and a plurality of them are related. The feature is that the viewing data of each individual is collectively transmitted.

A television camera is used as the viewer image input means, and a face image as identification information is cut out for each image frame from the video signal of the television camera.
By comparing and recognizing face images, a viewer is identified, viewer identification information is obtained every moment, and the identity of the viewer between image frames is determined. Also, a telephone line is used as a transmission means.

[0016]

According to the present invention, by adopting the above-mentioned structure, first, some image frames are sampled from the video signal of the viewer image input means (television camera or the like) to distinguish the person, and further, the specific viewing is performed. A face image as identification information is stored in the target from the person image of the person. Gender, age, occupation, etc. are input to the stored comparative identification information in correspondence with the face information. Then, by comparing and identifying the face image received from the image frame with the stored face image, the gender and age can be known from the identification information of the individual. In this way, the face images in the image frame are sequentially identified, and the personal identification information, the viewing time zone information, and the reception status information of the receiver at the same time are aggregated, thereby enabling a plurality of persons who are effective for the audience rating. It is possible to obtain the individual audience rating data.

In the determination of the personal identification information, if the unidentified face image still does not match the stored face image, it is stored as a person with an identification number as it is and transmitted as undetermined personal identification information. . The unidentified personal identification information is the same as the past when the facial image obtained from the extracted image frame is identified as the facial image of the particular viewer who is surely stored before or after the transmission. The data of the person is also ported together. Since the personal identification information is supplemented in this manner, accurate personal data at the time of viewing can be obtained.

Further, when a specific individual cannot be identified in image frame units, the determination is reserved, and the identity of the viewer among a plurality of image frames is used to reliably identify the individual. That is, it is possible to compare with face information that can be reliably determined, among a plurality of face information of the same undetermined person from the past to the present. The optimum face information is often an image frame in which the viewer is facing the screen.

Further, since the time calibration is carried out from the information processing device side, there is no time lag between the channel information and the personal time zone information, and the watching time can be accurately known.
Moreover, since the entire room is photographed and monitored, it is possible to grasp the number of people and the viewing attitude. Also, if multiple viewers are using several receivers, by organizing and sending personal time zone information and channel information in a batch,
Information processing can be performed promptly.

[0020]

Embodiments of the present invention will be described below with reference to the accompanying drawings. As shown in Figure 1, the structure of the survey system is
It is composed of 11 circuit blocks including a television (television receiver) 1. First, the television 1 has a function of outputting power on / off status and currently selected channel information in addition to a function of outputting normal video and audio, and transmits this signal to the central information processing apparatus A. It is connected so as to output to the communication device 2. In addition, a television camera (hereinafter, referred to as a camera) 3 having a viewing space for viewing the screen of the television 1 is installed on the television 1. With respect to the direction of the lens of the camera 3, it is preferable that the viewer's face be the front image when the viewer is looking directly at the television 1.

The video signal of the camera 3 is input to the frame memory 4. The video signal is moving image information in the room, and an image for one frame is taken out at an appropriate sampling interval and stored in the frame memory 4 in frame units. Although the sampling interval is an integral multiple of the frame period of the camera 3, it should be as short as possible for the purpose of pursuing audience rating measurement accuracy, and as long as possible from the viewpoint of processing speed. In fact, a time interval of 15 seconds to 1 minute is a reasonable design goal as a compromise between the above two requirements. As will be described later, it is possible to change the sampling interval to be short for the purpose of individual tracking and to be long for the purpose of individual recognition.

The cutout unit 5 detects a human image from the current image frame stored in the frame memory 4, cuts out the portion, and erases the other portions and sends them to the memory 6. In order to detect the person image, a known partial technique may be used to detect a moving partial image in the image frame and to detect a face image portion. When there are a plurality of persons in the current image frame, the images of the persons are stored in the memory 6.

The data arrangement inside the memory 6 is shown in FIG. Although this design example is designed to store the person images for three persons, the number of persons can be easily changed. In FIG. 3, an address 200 stores the number N of persons cut out in the current image frame (N = <<
3). Each of 201 to 205 stores a pointer P1 indicating the start address of the first person image, a horizontal size W1 of the image, a vertical size H1, and center coordinates X1 and Y1 of the person image. When the head address P1 is 1000 bytes, the horizontal size W1 is 40 (pixels), and the vertical size H1 is 60 (pixels), and the grayscale value of one pixel of the image is represented by 1 byte, the image of the first person is 1000- It is stored in 2400 bytes at address 3399. However, the unit of address is byte (8 bits).

FIG. 4 is a schematic diagram showing this person image.
206 to 210 include a pointer P2 indicating the start address of the image of the second person, a horizontal size W2 of the image, and a vertical size H.
2. The center coordinates X2 and Y2 of the person image are stored. At this time, P2 is address 3400 and W
When 2 = 16 and H2 = 25, the image of the second person is stored in 400 bytes at addresses 3400 to 3799. Numerals 211 to 215 denote a pointer P3 indicating the start address of the image of the third person, a horizontal size W3 of the image, and a vertical size H.
3, center coordinates X3, Y3 are stored respectively. At this time, P3 is address 3800, W3 = 40, H3
= 30, the images of the third person are 3800 to 4
It is stored in 1200 bytes at address 999.

The buffer memory 7 stores the person images in the past frames and has the same structure as the memory 6, but the number of person images to be stored may be different. In addition, the image size of each corresponding person is stored in the memory 6
And the buffer memory 7 may be different.

The face recognition section 8 assigns an identification number to the face image of each person in the input memory 6 once. The given face image is compared with the face image of the person registered in advance, and the ID information (personal identification information) of the most similar person is output. The ID information is provided corresponding to the registered person. In addition, ID information Z ₁ , Z ₂ , ... Representing unknown, which is output when a sufficiently similar person cannot be determined.
・ Is provided. Here, Z _n represents the n-th unknown person. In addition, not only the ID information of the most similar person but also the ID information of the second or lower candidate can be output to the person image of each person. The ID information output from the face recognition unit 8 is sent to the comprehensive determination unit 9.

The personal tracking unit 10 compares the image of each person in the memory 6 with the image of each person in the buffer memory 7,
It determines which partial image corresponds. For the determination, the coordinates of the center of the partial image are used as a basis, but the similarity of the images can be taken into consideration.

The comprehensive determination unit 9 uses the correspondence between the ID information output from the face recognition unit and the partial image determined by the personal tracking unit 10 to perform final personal recognition. That is, when the face image of the person Z _i determined to be unknown in the current image frame is located at a position continuous with the person images that appear in the past image frames of the face image determined to be the person A, the person Z _i is determined to be the person A. Further, when there is a face image of the person Z _j determined to be unknown in the past image frame, and this is at a position continuous with the person image i of the face image of the current image frame determined to be the person B , It is determined retroactively that the person Z _j is the person B in the past image frame. As a result, the ID information Z ₁ , which represents the unknown,
Z ₂ becomes small, and data is added to the already determined ID information, so that the information becomes accurate.

The ID information of the person judged by the comprehensive judgment unit 9 is transmitted from the communication device 2 to the central information processing device A in consideration of the viewing time zone. The communication device 2 may collectively output the stored ID information to the central information processing device A on the receiving side via the communication line 11 at an appropriate time, for example, at midnight or early morning. In this case, the channel information also needs to be temporarily stored (storage unit 1a). Further, the processing of transmitting the determination result of the comprehensive determination unit 9 from the communication device 2 moment by moment and determining the unknown ID information may be left to the information processing device A on the receiving side. However, in that case, it is also necessary to send information indicating the identity of the person between the frames. Furthermore, it is also possible to incorporate channel information into personal time zone information and transmit it as one signal. A plurality of pieces of information thus sent can be processed by the information processing apparatus A to obtain detailed audience rating data.

Further, in the above description, the individual tracking in the individual tracking section 10 and the face recognition in the face recognition section 8 are performed for the frames sampled at the same time intervals, but the individual tracking is performed at the smallest possible time intervals. Alternatively, face recognition may be performed at relatively coarse time intervals. By doing so, individual tracking can be reliably performed, so even if face recognition is unclear, comprehensive judgment is required. It is possible to correct by the above, and the load on the face recognition section 8 can be reduced.

Next, details of the cutout portion 5 will be described with reference to FIG. Reference numeral 4 is the frame memory shown in FIG. 1, in which the current indoor image is stored. Reference numeral 51 is also a frame memory, the contents of which are copied from the frame memory 4 at constant time intervals T (for example, 1 second). That is, the frame memory 51 stores the indoor image at a time point T past.

The images of the frame memory 4 and the frame memory 52 are input to the difference calculation circuit 52, and the difference image is created. The difference calculation circuit 52 compares the two input images for each pixel, creates an image in which the difference value is the pixel value, and stores the image in the frame memory 53. In the difference image in the frame memory 53, the room background without motion is erased, and the contour of the moving person image has a large value. For example, as shown in FIG. 5, a contour A represents a person whose whole body is moving, and a contour B shows a case where only the face moves.

The contour detecting section 54 detects the contour section from the difference image in the frame memory 53, obtains the coordinates of a rectangle indicating the outer contour section of the person, and supplies it to the temporary memory 55. In general, there may be a plurality of persons, and the temporary memory 55 is provided with coordinates regarding a plurality of person outlines. Face image detector
At 56, the face image portion is extracted from the frame memory 4 using the coordinates given from the temporary memory 55. That is, the portion of the person image is extracted, the pixels of the gray value corresponding to the face portion are extracted from the portion, the rectangle surrounding the pixels is determined, and the coordinates of the face image are stored in the temporary memory 57. It goes without saying that, when the coordinates of a plurality of person images are given to the temporary memory 55, the coordinates of the face image are given to the temporary memory 57 for each of them.

The face image extraction unit 58 uses this coordinate information to extract the face image portion from the frame memory 4 and stores it in the memory 6. The structure of the memory 6 is as described above, but it is clear that the information required therein is obtained from the temporary memory 57,

Next, details of the face recognition section 8 will be described. The face recognition unit 8 recognizes the face image in the memory 6 at every suitable time interval T _R , but it is convenient to set the time interval T _{R to} an integral multiple of the image acquisition time interval T. . For example, T = 1 second and T _R = 15 seconds are practically suitable values. The face recognition unit 8 is provided with a plurality of face images stored in the memory 6, but since they are the same in principle, 1
The recognition of the face image of the name will be described.

As shown in FIG. 6, the face image for one person is copied from the memory 6 to the buffer memory 81 of the face recognition section 8. From this face image, the autocorrelation coefficient calculation circuit 82 calculates the autocorrelation coefficient of the face image and stores it in the buffer memory 83.

Here, the autocorrelation coefficient is Is the amount defined by. Here, P (i, j) indicates the value of (i, j) pixel of the image, and M and N indicate the size of the face image. Further, (ξ, η) represents the relative movement amount of the image. If calculation is performed in the range of −R = <ξ = <R, −R = <η = <R, K = (2R + 1) ² self The correlation coefficient is obtained. An array of these K autocorrelation coefficients is the feature vector of the face image to be personally recognized, that is, the unknown face image.

The autocorrelation coefficient (K-dimensional vector) stored in the buffer memory 83 is sequentially compared with the standard pattern of each individual stored in the memory 85 under the control of the control unit 84 (comparing unit). 86). The standard pattern is a feature vector (autocorrelation function, K-dimensional vector) calculated in advance for each person to be recognized. If the feature vector of the Kth viewer is σ _K (ξ, η), the distance between the unknown face image and the Kth viewer face image is Given in.

This distance is stored in the memory 87. After the distances for all the viewers to be recognized are stored in the memory 87, by the control of the control unit 84, m that gives the minimum distance is obtained and stored in the memory 88 as a recognition result. However, when D _m => δ, the recognition result is suspended and the answer is returned as unrecognizable. Here, δ is an appropriately determined parameter.

There may be a case where a plurality of unrecognizable faces are generated by the face recognition section 8. Generally, in order to distinguish this, Z ₁ ,
It is expressed as Z ₂ . For example, expressing Z _n = −n is one method. The recognition result is sent to the comprehensive judgment unit 9 in FIG. Even in the case of being unrecognizable, it is effective to send the identification number of the undetermined person to the comprehensive determination unit 9 in association with the identification number of the specific viewer who approximates the identification number. In this case, if the identification number of the particular viewer and m, to the Z _n = -m as a recognition result, also is another one of the representations.

Next, details of the personal tracking unit 10 will be described with reference to FIG. The memory 6 and the buffer memory 7 are those shown in FIG. 1, and as described above, the memory 6 stores the face images of the persons in the current image frame, and the buffer memory 7 stores the face images of the persons in the past image frame. The face image of is stored and is input to the comparison computer 10a. Since the internal structures of the memory 6 and the buffer memory 7 are as shown in FIG. 3, the central coordinates of each human face image can be obtained from this. In the comparison computer 10a, the center coordinates of the face images of the persons are compared, and the correspondence relationship between the face images of the persons in the memories 6 and 7 is obtained. That is, the center coordinates (xi
^When the distance between ( ⁶ , yi ⁶ ) and (xj ⁷ , yj ⁷ ) is small, it is determined that the person i in the memory 6 and the person j in the buffer memory 7 are the same. This judgment result is sent to the comprehensive judgment unit 9.

Next, the details of the comprehensive judgment unit 9 will be described with reference to FIG. The determination results of the face recognition unit 8 and the personal tracking unit 10 are given to the comprehensive determination unit 9. The recognition result of the face recognition unit 8 is stored in the memory 91. The result of the individual tracking unit 10 is given to the processing unit 92, and the processing unit 92 integrates the content of the memory 91 and the result of the individual tracking unit 10 to correct the content of the memory 91. The output of the contents of the memory 91 is the comprehensive determination result. The internal structure of the memory 91 is shown in FIG. In FIG. 9, each row shows the recognition result at a certain temporary point (corresponding to a unit image frame), and each column shows each individual image in the memory 6. It should be noted that A to D in FIG. 9 are for easy understanding of the passage of time.

In the first row T1 of FIG. 9A, the person 2 is recognized as the viewer 1 and the persons 1 and 3 are not recognized. That is, since the persons 1 and 3 were not identified as the specific viewers, the numerical values Z ₁ ¹ and Z ₂ ¹ indicating unrecognizable
Goes in. However, the superscript of Z is for distinguishing the time. In the second row T2, persons 1, 2, 3
For the specific viewers 2, 1, 3 respectively,
Is obtained as the output of. Although the persons 1, 2, and 3 may be moving in the image frame, here, the persons at consecutive positions in the image frame are arranged as the same person in the table of FIG.

At that time, in the processing unit 92, the persons 1, 2, 3
Replaces the viewer identified in the second row that matches the person in the first row T1. By tracing back and comparing the results of the second row T2 to the uncertain person in the first row T1, Z ₁ ¹ = 2 and Z ₂ ¹ = 3 can be confirmed. Therefore, it becomes like B of FIG.

For the third row T3, the unknown person Z
By transmitting the recognition result of the second row T2 to ₁ ³ (B in FIG. 9), Z ₁ ³ = 1 can be determined (C in FIG. 9). Is the same also for the fourth row T4, the unknown person Z ₂ ^4, remain as unknown because historical information is not available (D in FIG. 9), and waits for subsequent processing.

Next, another embodiment will be described with reference to FIG. Here, a people counter 13 is provided. The number of people counter 13 counts the number of people in the room,
For example, an ultrasonic sensor can be used. The comprehensive determination unit 9 compares the number-of-persons information with the number of persons obtained as the output of the face recognition unit 8 and, if they do not match, feeds them back to the cutout unit 5 to re-perform recognition. good.

Further, the face recognition section 8 identifies the direction in which the face is facing and sends it to the communication device 2 as a part of the ID information. This information is used as information indicating a television viewing attitude. Therefore, the central information processing device A that has received this information can create more detailed audience rating data, which can be useful for improving the quality of television programs.

A calibration circuit for the timepiece 12 may be provided,
In this case, the communication device 2 receives the time calibration signal via the communication line 11 and activates the calibration circuit. For example, by sending a signal every day at 12:00 pm, it is possible to avoid the generation of erroneous data due to the accumulation of time-dependent errors in the clock.

Further, in a home having a plurality of televisions 1, this system may be attached to each room in which each television 1 is installed and the communication device 2 may be shared. In this case, since communication of one system is stopped, a storage device between them is provided. In addition, even if there is a plurality of televisions 1, when only one television 1 is in operation at a time, there is always one channel information, so there is a method in which the data of a plurality of people is put on one channel information. In addition to this, there is a system in which channel information of a plurality of televisions 1 is added to personal time zone information of one individual for communication. As a result, it is possible to accurately obtain the audience rating data at home regardless of which television receiver the individual uses.

In the above embodiment, it has been assumed that the standard pattern of the face registered by each individual has already been registered in the face identification section 8. The learning method of the standard pattern will be described below.

First, several designated viewers stand in front of the camera 3, and the faces facing the front and the respective directions are displayed without moving each time, and an image is taken by the camera 3. Image frames are continuously sampled from the output of the camera 3, and each face image is extracted from the image frame by the cutout unit 5 to create data, which is registered in the face recognition unit 8 as standard pattern information for individual identification, and ID information is registered. To be output as. When each face image is extracted, each viewer is given a switch that allows each individual to be distinguished, and when this switch is pressed, the image is cut out, and at the same time the switch distinguishes the gender, age, and occupation. It is possible to add information such as. It is better not to learn once but to do it several times.

The standard patterns thus collected may be one pattern on average, but are registered as separate patterns. In addition, the same ID information may be output even if it matches any of the registered patterns. In addition, each person is provided with a position information transmission device such as an ultrasonic transmitter, and using this position information, the location of the individual on the screen is estimated, and the personal image is cut out in the image frame. By doing so, it is possible to prevent the cutting error from occurring at the time of learning, and it is possible to prevent the false recognition caused by the learning error.

As described above, the camera 3 captures an unspecified number of viewers who are actually watching. The person captured by the camera 3 is tracked as a moving image, and a specific individual is identified by a face image. When confirmed, the past viewing data of the person is also added. Therefore, finally, the unknown person becomes a person other than the audience rating cooperator, and the numerical value of the data can be made reliable. Furthermore, since it is possible to know what kind of posture the viewer is looking at, what time it is, it is possible to automatically obtain the viewer rating data for each individual, and without burdening the viewer. In addition, it is not necessary to request operation of a switch or the like at the start and end of viewing, and it is possible to avoid a decrease in accuracy due to forgetting to press the switch.

[0054]

Since the present invention is configured as described above, the identification information of the viewer is extracted from the image frame and compared with the identification information prepared in advance to identify the specific viewer, Furthermore, since the viewing data in the past is used as a reference, for those who cannot be identified, the data is accumulated by the identification number and converted to the personal identification information at the time of establishment, so highly accurate personal identification is performed. , It is possible to acquire highly reliable individual audience rating data. Further, as a result, the reaction of the viewer can be accurately grasped, which can contribute to the improvement of the level of television broadcasting. Further, by using a television camera as the viewer image input means, it is possible to collect high-quality individual audience rating data including not only the presence of the viewer but also the viewing attitude of each viewer.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an audience rating survey system according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a detailed configuration of a cutout portion shown in FIG.

FIG. 3 is an explanatory diagram showing a data arrangement inside the memory shown in FIG. 1.

FIG. 4 is a schematic diagram of a face image based on the data in the memory shown in FIG. 1.

5 is a schematic diagram of a difference image represented by data inside the memory 43 shown in FIG.

FIG. 6 is a block diagram showing a detailed configuration of a face recognition section shown in FIG. 1.

7 is a block diagram showing an individual tracking unit in the configuration of the audience rating survey system shown in FIG. 1. FIG.

FIG. 8 is a block diagram showing details of a comprehensive determination unit shown in FIG.

9 is an explanatory diagram showing an internal data arrangement of the comprehensive determination unit shown in FIG.

FIG. 10 is a block diagram showing the configuration of an audience rating survey system according to another embodiment.

[Explanation of symbols]

 1 TV receiver 3 TV camera 8 Face recognition unit 9 Comprehensive judgment unit 10 Personal tracking unit 12 Clock

Claims (9)

[Claims] 1. A television receiver, a viewer image input means directed to a viewing space of the television receiver, and viewer identification information cut out from the image frame extracted by the viewer image input means. Obtaining means, means for pre-storing identification information of a specific viewer, identification information of the viewer newly extracted by the viewer image input means, and identification information of the specific viewer stored by the previous storage means. And a clock means for detecting the time zone of the input signal of the newly extracted viewer identification information added to the determination means, and the personal time zone information obtained by the determination means. And a transmission means for transmitting the channel information of the television receiver to an information processing device, the audience rating survey system. 2. When a plurality of extracted identification information are confirmed at consecutive positions among a plurality of consecutively extracted image frames, the same identification number is assigned to the consecutive plurality of identification information and stored. Storage means, determination means for identifying a new person who has been given an identification number and a specific viewer, and replacing the identification number or stored information of the person who has been given the identification number as identification information for the specified viewer. The audience rating survey system according to claim 1, further comprising: 3. The determination means for sending the identification information of the unidentified identification number to the information processing apparatus, and the transmission when the person who gives the unidentified identification number is determined to be the same as the identified viewer. 2. The audience rating survey system according to claim 1, further comprising means for transmitting a signal for associating the personal time zone information with the subsequently identified personal time zone information. 4. Upon receiving a signal from a side performing information processing,
The audience rating survey system according to claim 1, further comprising means for calibrating the time of the clock means. 5. The audience rating survey system according to claim 1, wherein the number of persons existing in the viewing range is counted from the image frame, and the number information is transmitted. 6. A mode of viewing is identified from the image frame,
The audience rating survey system according to claim 1, wherein the information indicating the viewing mode is transmitted. 7. The identification information of a specific person is assigned an identification number to be learned and stored, and the identification information assigned the identification number is replaced with the identification information of a specific viewer. The audience rating survey system according to claim 1. 8. The audience rating survey according to claim 7, further comprising a position confirmation device for each person, wherein the image of each person in the screen frame is identified from the output of the position confirmation device during learning. system. 9. A plurality of systems according to claim 1 are installed,
An audience rating survey system characterized by comprehensively determining personal time zone information and channel information for a plurality of television receivers, and collectively transmitting viewing data of each individual regarding a plurality of television receivers.