WO2012020591A1

WO2012020591A1 - System for identifying individuals, feature value specification device, feature specification method, and recording medium

Info

Publication number: WO2012020591A1
Application number: PCT/JP2011/062313
Authority: WO
Inventors: 昭裕早坂
Original assignee: 日本電気株式会社
Priority date: 2010-08-09
Filing date: 2011-05-24
Publication date: 2012-02-16
Also published as: JPWO2012020591A1

Abstract

Appropriate feature values for use in the identification of individuals are specified for each of a plurality of identification targets. A feature value specification device is provided with a registration unit, an effective value calculation unit, and feature value specification unit. The registration unit receives: information indicating an individual that is an identification target and is detected from environment information originating in the spatial environment where the individual is present; position information indicating the position of each individual; feature values extracted from the environment information; and attribute information indicating the attributes of each feature value. For each feature value, the registration unit uses the position information and the attribute information to determine the individual that generated the feature value. For every feature value extracted from the environment information, the effective value calculation unit obtains an effective value indicating the quality of that feature value. For each individual detected, the feature value specification unit uses the determination result and the effective value to specify each feature value used to identify that individual.

Description

Individual identification system, feature amount specifying device, feature amount specifying method, and recording medium

The present invention relates to an individual identification system, a feature quantity specifying device, a feature quantity specifying method, and a recording medium, and in particular, an individual identification system, a feature quantity specifying device, and a feature quantity for specifying a feature quantity for each of a plurality of individuals to be identified. The present invention relates to a specifying method and a recording medium.

There is known a technique for performing authentication using a plurality of feature amounts when identifying an individual. In a technique for performing individual identification using a plurality of feature amounts, the quality of a feature detection signal used for collation for individual identification is not considered. Therefore, for example, in a situation where the measurement conditions of features are not good and the signal-to-noise ratio (signal-to-noise ratio) is bad, the error occurrence rate in matching using all the features uses only features with good detection signal quality. It is higher than the error occurrence rate in collation.
An example of a technique for solving this problem is described in Patent Document 1.
The personal recognition system described in Patent Literature 1 includes an acquisition unit, a collection unit, a calculation unit, a specifying unit, and a recognition unit. The acquisition unit acquires environmental information of a space where a person exists. The collecting unit collects a plurality of types of feature amounts from the environment information acquired by the acquiring unit. The calculating means obtains an effective value for each feature amount collected by the collecting means. The specifying unit specifies a feature amount used for the personal recognition process based on each effective value calculated by the calculating unit. The recognition means performs personal recognition processing using the feature amount specified by the specifying means.
The effective value is a numerical value representing the quality of the feature detection signal. For example, the effective value is an index that enables a process of selectively using features with better detection signal quality for matching when the feature measurement conditions are not good and the SN ratio is poor.
The personal recognition system having such a configuration operates as follows.
That is, the acquisition unit acquires the environment information, and the collection unit collects a plurality of types of feature amounts from the acquired environment information. Then, the calculating means calculates the effective value of each feature, and the specifying means specifies the feature amount used for the personal recognition processing based on the calculated effective value. The recognizing means performs personal recognition processing using the feature amount specified by the specifying means.
Patent Document 2 describes a technique for simultaneously authenticating a plurality of individuals.
The authentication system described in Patent Literature 2 includes a weight measurement floor configured by a plurality of weight measurement units, a weight sensor unit, a controller, and an information processing device. When a plurality of authentication subjects pass through the weight measurement floor of this authentication system, each weight sensor unit measures the weight. The measurement result is notified to the information processing apparatus via the controller. Then, the information processing apparatus separates information about each authentication target person and obtains walking data such as movement of the center of gravity. Then, the information processing apparatus compares the stored data with each other, and if they match, gives an authentication permission to the person to be authenticated.
Patent Document 3 describes a technique for associating each of a plurality of feature amounts output from a plurality of individuals with an individual that has output each feature amount.
The conference system described in Patent Document 3 includes a plurality of microphones, voice recognition means, position specifying means, association means, and synthesis means. The voice recognition means recognizes the voice input to each microphone. Further, the position specifying means specifies the position of the speaker in the captured image. The associating means associates the remark character information obtained by the recognition by the voice recognition means and the user position information indicating the position of the speaker specified by the position specifying means. The synthesizing unit synthesizes the character image related to the comment character information into an image portion corresponding to the position indicated by the user position information in the captured image.

JP 2006-293644 A Japanese Patent Laying-Open No. 2005-078228 JP 2009-194857 A

The personal recognition system described in Patent Document 1 described above assumes only the case where the feature amount included in the environment information acquired by the acquisition unit is generated from one recognition target. Therefore, when there are a plurality of recognition targets in the space for acquiring environment information, the personal recognition system described in Patent Document 1 is a feature amount in which all the feature amounts collected by the collecting means are obtained from one target. It is considered that. Therefore, it is difficult for the personal recognition system described in Patent Document 1 to specify an appropriate feature amount for use in individual recognition for each of a plurality of recognition targets.
For example, there are two persons in the space for acquiring environment information, only the facial feature quantity is acquired from one person, and only the audio feature quantity is acquired and acquired from the other person. In addition, the effective value of each feature amount may be high. In this case, the personal recognition system described in Patent Document 1 determines that the acquired facial feature quantity and voice feature quantity are information obtained from the same person. That is, the personal recognition system described in Patent Document 1 results in erroneous identification.
The authentication system described in Patent Document 2 described above performs authentication based on one type of feature amount. Therefore, in the authentication system described in Patent Document 2, when a plurality of feature amounts are collected, it is difficult to specify an appropriate feature amount for use in individual authentication for each of a plurality of authentication targets. is there.
The conference system described in Patent Document 3 described above associates each feature quantity based on the position where the feature quantity obtained from the image information is detected and the position where the feature quantity obtained from the audio information is detected. The conference system identifies the individual by performing individual recognition based on the feature amount obtained from the image information. Then, the conference system associates the feature quantity obtained from the individual and the image information with the feature quantity obtained from the audio information. That is, since the conference system described in Patent Document 3 cannot identify an individual based on a plurality of feature amounts, it is necessary to identify an appropriate feature amount for use in individual identification for each of a plurality of identification targets. It is difficult.
An example of an object of the present invention is to provide an individual identification system, a feature amount specifying device, a feature amount specifying method, and a recording medium for specifying an appropriate feature amount for use in individual identification for each of a plurality of identification objects. It is in.

The first feature amount specifying device according to one aspect of the present invention includes information indicating an individual detected from environment information that is information derived from an environment in a space where an individual to be subjected to individual identification processing exists, and the individual Each of the position information indicating the position, the feature amount extracted from the environment information, and the attribute information indicating each attribute of the feature amount, respectively, and based on the position information and the attribute information, For each feature quantity, a registration unit that discriminates the individual in which the feature quantity has occurred; for each feature quantity extracted from the environment information, an effective value calculation unit that obtains an effective value indicating the quality of the feature quantity; and And a feature amount specifying unit that specifies a feature amount used for individual identification processing of the individual for each detected individual based on the determination result and the effective value.
The first individual identification system according to one aspect of the present invention includes an environmental information acquisition unit that acquires environmental information that is information resulting from an environment of a space where a plurality of individuals to be subjected to individual identification processing exist, and the environmental information An individual detection unit for detecting the plurality of individuals together with position information indicating the position of each individual, and feature amount extraction for extracting a plurality of feature amounts from the environment information together with attribute information indicating the type of each feature amount A registration unit that determines, for each feature quantity, an individual in which the feature quantity has occurred, based on the position information of each of the individuals and the attribute information of the feature quantities, and the environment information For each extracted feature quantity, an effective value calculation unit for obtaining an effective value indicating the quality of the feature quantity, and for each detected individual based on the result of the discrimination and the effective value. It includes a feature amount specifying unit for specifying a feature quantity for use in individual identification process of an individual, and a matching unit for performing individual identification process for each individual that has been extracted on the basis of the identified feature value.
The first feature amount specifying method according to one aspect of the present invention includes: information indicating an individual detected from environment information that is information derived from an environment in a space where an individual that is an object of individual identification processing is present; and the individual Each of the position information indicating the position, the feature amount extracted from the environment information, and the attribute information indicating each attribute of the feature amount, respectively, and based on the position information and the attribute information, For each feature amount, an individual in which the feature amount has occurred is determined, and for each feature amount extracted from the environment information, an effective value indicating the quality of the feature amount is obtained, and the determination result and the effective value are obtained. Based on the detected individual, the feature quantity used for the individual identification process of the individual is specified.
The first recording medium in one aspect of the present invention includes information indicating the individual detected from environmental information that is information derived from the environment of the space where the individual that is the target of the individual identification process exists, and each of the individual Receiving the position information indicating the position of the feature, the feature quantity extracted from the environment information, and the attribute information indicating the attribute of the feature quantity, respectively, and based on the position information and the attribute information, the feature quantity A process for discriminating an individual in which the feature quantity has occurred, a process for obtaining an effective value indicating the quality of the feature quantity for each feature quantity extracted from the environment information, a result of the discrimination and the effective value For each detected individual, a program for causing the computer to execute a process for specifying a feature amount used for the individual identification process of the individual is recorded.

One of the effects of the present invention is that it is possible to specify an appropriate feature amount for use in individual identification for each of a plurality of identification targets.

FIG. 1 is a block diagram illustrating a configuration example of an individual identification system 1 according to the first embodiment. FIG. 2 is a block diagram illustrating a configuration example of the sensor unit 1000 according to the first embodiment. FIG. 3 is a block diagram illustrating a configuration example of the feature amount determination unit 1103 according to the first embodiment. FIG. 4 is a flowchart illustrating an operation example of the individual identification system 1 according to the first embodiment. FIG. 5 is a block diagram illustrating a configuration example of the individual identification system 2 according to the second embodiment. FIG. 6 is a block diagram illustrating a configuration example of the composite sensor unit 2200 according to the second embodiment. FIG. 7 is a flowchart illustrating an operation example of the individual identification system 2 according to the second embodiment. FIG. 8 is a block diagram illustrating a configuration example of the individual identification system 3 according to the third embodiment. FIG. 9 is a block diagram illustrating a configuration example of the biological information acquisition unit 3200 according to the third embodiment. FIG. 10 is a flowchart illustrating an operation example of the individual identification system 3 according to the third embodiment. FIG. 11 is a block diagram illustrating a configuration example of the individual identification system 4 according to the fourth embodiment. FIG. 12 is a flowchart illustrating an operation example of the individual identification system 4 according to the fourth embodiment. FIG. 13 is a block diagram illustrating a configuration example of the feature quantity specifying device 50 according to the fifth embodiment. FIG. 14 is a diagram illustrating an example of information stored in the database unit 1107.

Next, embodiments for carrying out the invention will be described in detail with reference to the drawings.
[First Embodiment]
FIG. 1 is a block diagram showing a configuration of an individual identification system 1 according to the first embodiment of the present invention.
As shown in FIG. 1, the individual identification system 1 according to the first embodiment of the present invention includes a sensor unit 1000 and an identification processing device 1100.
The configuration of the sensor unit 1000 will be described.
The sensor unit 1000 acquires environmental information that is information resulting from the environment of the space in which the individual that is the target of the individual identification process exists. Hereinafter, a space in which an individual subject to individual identification processing exists is simply referred to as a real space.
The environmental information may include image information in real space. In this case, as a specific configuration, the sensor unit 1000 may include an image information acquisition unit 1001 that acquires real space image information as illustrated in FIG. 2. The environment information may include audio information in real space. In this case, as a specific configuration, the sensor unit 1000 may include an audio information acquisition unit 1002 that acquires audio information in real space as illustrated in FIG.
Specifically, the image information acquisition unit 1001 may be a video camera capable of capturing a real space image or a still image. Further, the sound information acquisition unit 1002 may be a microphone that can acquire sound in real space. The microphone constituting the audio information acquisition unit 1002 may have a directivity function that can specify the position where the audio is generated.
Note that a plurality of the image information acquisition unit 1001 and the audio information acquisition unit 1002 may be arranged at two or more positions in the real space.
In addition, the means for acquiring environment information of the real space is not limited to the above configuration. For example, the means for acquiring the real space environment information may be acquired by reading the real space environment information held in the external storage device.
In the present embodiment, it is assumed that the environment information is information including image information and audio information.
The configuration of the identification processing device 1100 will be described.
The identification processing device 1100 includes an individual detection unit 1101, a feature amount extraction unit 1102, a feature amount determination unit 1103, a feature amount identification unit 1104, a database addition processing unit 1105, a collation unit 1106, and a database unit 1107.
The identification processing device 1100 according to the first embodiment determines an individual in which each feature amount extracted from the environment information has occurred. Then, the identification processing device 1100 obtains an effective value indicating the quality of the feature value for each feature value extracted from the environment information. Then, the identification processing device 1100 specifies a feature amount used for the individual identification process for each individual detected from the environment information based on the determined result and the calculated effective value.
Therefore, the identification processing device 1100 according to the first embodiment uses a feature value having a high effective value, that is, a higher quality feature among the feature values associated with each individual, and uses the feature value for individual identification processing. Can do.
Hereinafter, each component provided in the identification processing device 1100 according to the first embodiment will be described in detail.
The individual detection unit 1101 receives environmental information of the real space acquired by the sensor unit 1000 from the sensor unit 1000. Then, the individual detection unit 1101 performs a process of detecting an individual to be identified that exists in the real space from the received environment information. The individual detection unit 1101 also identifies the position of the individual together with the individual from the received environment information.
Specifically, the individual detection unit 1101 may specify the individual to be identified and the position of the individual by applying a background difference method or pattern matching to the image information included in the acquired environment information. . Alternatively, when the individual detection unit 1101 receives information including a multi-viewpoint image as environment information, the individual detection unit 1101 may specify an individual to be identified and a three-dimensional position in the space of the individual using the multi-viewpoint image. Or the individual detection part 1101 may detect the individual used as identification object as follows, and may specify the position of the individual. For example, when the sensor unit 1000 includes a directional microphone, the individual detection unit 1101 may specify the position of the pronunciation individual by the following method. That is, the individual detection unit 1101 detects a sounding individual based on the environmental information received from the sensor unit 1000 and the estimated position of the sounding individual specified by the function of the directional microphone included in the sensor unit 1000, and determines the position of the sounding individual. You may specify.
In addition, the individual detection unit 1101 generates position information indicating the position of the specified individual for each detected individual.
The individual detection unit 1101 includes, for each detected individual, a feature quantity extraction unit 1102 that includes information indicating the individual, position information of the individual, and image information and audio information of at least a part of the environment information corresponding to the position information. Output to. At least a part of the image information may be image information including a region of the individual. The image information may be video information including the area of the individual. Further, at least a part of the voice information may be voice information indicating a voice estimated to be generated within a predetermined distance from the position indicated by the position information of the individual.
In addition, the individual detection unit 1101 outputs, for each detected individual, information indicating the individual and position information of the individual to a feature amount specifying unit 1104 described later.
The feature amount extraction unit 1102 receives information indicating an individual, position information of the individual, image information corresponding to the individual, and audio information from the individual detection unit 1101. The feature amount extraction unit 1102 extracts feature amounts from the received image information and audio information. Specifically, the feature amount extraction unit 1102 may extract feature amounts based on the color, shape, size, pattern, etc. of the object that can be acquired from the received image information. Alternatively, the feature amount extraction unit 1102 may extract a feature amount based on an action of an object that can be acquired from the received video information. Alternatively, the feature amount extraction unit 1102 may extract a feature amount based on sound emitted from an object that can be acquired from the received audio information.
The feature quantity extraction unit 1102 specifies the attribute of the feature quantity when extracting the feature quantity. The attribute of the feature amount may be, for example, the following attribute.
(A) Type of information from which feature quantity is acquired
(B) Whether or not the feature quantity originated from a human
(C) Whether the feature quantity was generated from a man or a woman
(D) The race of the person who generated the feature
(E) Feature quantity strength
(F) Feature level
(G) Whether the feature quantity is a language
(H) How many words the feature value is
(I) Position where the feature amount is estimated to have occurred
The feature amount extraction unit 1102 generates attribute information indicating the attribute of the specified feature amount. For example, if the attribute of the feature value is (a) “type of information from which the feature value is acquired”, the feature value extraction unit 1102 determines whether the feature value is obtained from audio information in the environment information, or Specify whether it was obtained from image information. Then, the feature amount extraction unit 1102 generates information indicating the specified acquisition source as attribute information.
For example, if the attribute of the feature quantity is (b) “whether or not the feature quantity originated from a person”, the feature quantity extraction unit 1102 analyzes the feature quantity and includes information peculiar to the person. Judge whether or not. Any known method can be applied as this determination method. When the feature quantity includes information unique to a person, the feature quantity extraction unit 1102 generates information indicating that the feature quantity is generated from a person as attribute information.
For example, if the attribute of the feature value is (c) “whether the feature value is generated from a man or a woman”, the feature value extraction unit 1102 analyzes the feature value and determines whether the feature value is generated from a man or a woman. Judgment is made from Any known method can be applied as this determination method. When the feature quantity includes information specific to men or women, the feature quantity extraction unit 1102 generates, as attribute information, information indicating that the feature quantity is generated from a man or a woman.
Even if the attribute of the feature quantity is other than the above (a) to (c), the feature quantity extraction unit 1102 generates predetermined attribute information in the same manner as described above.
The feature quantity extraction unit 1102 outputs the extracted feature quantity and attribute information of each feature quantity to a feature quantity determination unit 1103 described later.
The feature quantity discriminating unit 1103 discriminates from which individual in the real space the feature quantity extracted by the feature quantity extracting unit 1102 is a feature quantity, and outputs a discrimination result. Then, the feature amount determination unit 1103 calculates an effective value of each feature amount. Specifically, as illustrated in FIG. 3, the feature amount determination unit 1103 includes a registration unit 1113 that associates an individual with a feature amount, and an effective value calculation unit 1123 that calculates an effective value of each feature amount.
The registration unit 1113 receives information indicating an individual and position information of each individual from the individual detection unit 1101. The registration unit 1113 receives the feature amount and attribute information of each feature amount from the feature amount extraction unit 1102. Then, the registration unit 1113 determines an individual in which each of the received feature amounts has occurred based on the received position information and attribute information.
Specifically, the registration unit 1113 may associate the individual and the feature amount by the following method. First, the registration unit 1113 specifies the difference between the position of each individual specified by the individual detection unit 1101 based on the environment information and the position where each feature amount specified by the feature amount extraction unit 1102 is estimated to have occurred. . The registration unit 1113 associates the individual with the feature amount when the identified difference is equal to or less than a predetermined threshold.
Alternatively, the registration unit 1113 may associate an individual with a feature amount by the following method. That is, the registration unit 1113 associates the individual with the feature quantity when the difference in time of the position of each individual corresponds to the difference with time of the position where each feature quantity is estimated to occur. . Here, “corresponding” may mean a case where each difference is a vector and the vector specified by the difference between the vectors is less than a predetermined length.
In this case, the individual detection unit 1101 and the feature amount extraction unit 1102 may associate information indicating the position of the individual or the time when the attribute of the feature amount is specified with information or feature amount indicating each individual. Then, the individual detection unit 1101 and the feature amount extraction unit 1102 may output information or feature amount indicating an individual associated with information indicating time to the feature amount determination unit 1103.
Alternatively, the sensor unit 1000 may associate the time when the environment information is acquired with the environment information, and pass the environment information to the individual detection unit 1101 or the feature amount extraction unit 1102. The individual detection unit 1101 or the feature amount extraction unit 1102 may associate the time associated with the received environment information with the position information indicating the position of the individual or the attribute information indicating the attribute of the feature amount. Then, the individual detection unit 1101 or the feature amount extraction unit 1102 may output the position information of the individual associated with the time or the attribute information of the feature amount to the feature amount determination unit 1103.
Alternatively, the registration unit 1113 may associate an individual with a feature amount by the following method. When the attribute of the feature quantity is (b) “whether or not the feature quantity originated from a human being” and the attribute information indicates information “human being”, the registration unit 1113 uses the following method to Correspond with the feature value. That is, the registration unit 1113 may perform speaker estimation based on the feature amount indicating the movement of the lips among the feature amounts included in the image information. Then, the registration unit 1113 associates the feature amount extracted from the image information estimated to indicate the speaker with the feature amount extracted from the speech information. Finally, the registration unit 1113 may specify an individual to be associated with the above-described associated feature quantity based on the relationship between each feature quantity and the individual. As a result, the registration unit 1113 can perform the association with higher accuracy than the association based on the one-to-one relationship between the feature amount and the individual.
Note that the registration unit 1113 may assign one feature amount to a plurality of individuals by probabilistic weighting in addition to assigning one feature amount to only one individual. For example, the registration unit 1113 associates one feature amount with a weight of a probability of 80% for the first individual and associates it with a weight of the probability of 20% for the second individual. Good.
This probability is a probability indicating the likelihood that the feature amount has occurred from the individual. This probability may be calculated by the following method. For example, when the feature quantity attribute information includes position information indicating the position where the feature quantity is estimated to occur, the registration unit 1113 is specified by the position information of each individual and the attribute information of each feature quantity. The above-described probability may be calculated based on the difference between the positions. For example, the registration unit 1113 may assign a higher probability as the difference is smaller and assign a lower probability as the difference is larger. Specifically, the registration unit 1113 may assign probabilities so as to be inversely proportional to the difference.
Further, this probability may be determined based on the feature amount attribute information. For example, when the feature quantity attribute information includes information indicating the type of feature quantity, the registration unit 1113 may weight each of the above probabilities according to the type of feature quantity.
The effective value calculation unit 1123 calculates an effective value indicating the quality of the feature amount based on the signal intensity of the image information and the sound information from which the feature amount is extracted. As an effective value calculation method, there are a method of calculating quantitatively using information of signal SN ratio and signal intensity, and a method of calculating by a neural network or support vector machine using previously collected learning data. Further, as a method for calculating the effective value, the method described in Patent Document 1 may be applied. These are examples, and the calculation method of the effective value is not limited to the above.
The feature amount specifying unit 1104 specifies a feature amount used for identifying an individual for each individual detected by the individual detecting unit 1101 from the feature amounts extracted by the feature amount extracting unit 1102. Specifically, the feature quantity specifying unit 1104 uses the feature quantity discrimination result and effective value obtained by the feature quantity discrimination unit 1103 to specify the feature quantity used for collation for individual identification processing.
For example, the feature quantity specifying unit 1104 refers to the effective value of the feature quantity determined to be generated from one individual for each individual detected by the individual detection unit 1101. Then, the feature amount specifying unit 1104 specifies a feature amount whose effective value referred to is equal to or greater than a predetermined threshold as a feature amount to be used for individual identification processing of the individual. This predetermined threshold may be a predetermined constant. Alternatively, the predetermined threshold may be a value calculated by the feature amount specifying unit 1104 based on the feature amount attribute information.
When the registration unit 1113 assigns one feature amount to a plurality of individuals with weighting probabilistically, the feature amount specifying unit 1104 detects the feature amount used for individual identification by the following method. It may be specified for each individual detected by the unit 1101. That is, for each individual detected by the individual detection unit 1101, the feature amount specifying unit 1104 is estimated that the effective value of the feature amount determined to have occurred from one individual and that the feature amount has been generated from that one individual. Each product with the probability is calculated. Then, when the calculated product is equal to or greater than a predetermined threshold, the feature amount specifying unit 1104 specifies the feature amount as a feature amount used for the individual identification process of the one individual.
The feature quantity specifying unit 1104 may specify a feature quantity used for individual identification processing of each individual by referring to a database unit 1107 described later. Specifically, the feature quantity specifying unit 1104 determines whether or not a feature quantity associated with information indicating a certain individual is stored in the database unit 1107. When the feature amount specifying unit 1104 determines that the feature amount is stored in the database unit 1107, the feature amount having attribute information corresponding to the attribute information of the feature amount is used for the individual identification processing of the individual. Is identified.
The database unit 1107 stores a plurality of types of feature amounts acquired from various individuals. The type of feature amount includes an image feature based on a texture pattern such as a color, shape, size, and pattern of an object, a video feature based on the motion of the object, an audio feature based on sound generated by the object, and the like. Each type of feature amount is registered in the database unit 1107 in association with each individual.
Note that the database unit 1107 may be configured to be installed outside the identification processing device 1100 as long as it is communicably connected to the identification processing device 1100. Each feature amount may be stored in the database unit 1107 in association with information indicating the type of the feature amount. Further, each feature amount may be stored in the database unit 1107 in association with the attribute information of the feature amount. Further, each feature amount may be stored in the database unit 1107 in association with information indicating an individual associated with the feature amount by the feature amount determination unit 1103.
FIG. 14 is an example showing information stored in the database unit 1107. Referring to FIG. 14, the database unit 1107 stores information indicating an individual, a feature amount associated with each individual, and attribute information of the feature amount in association with each other. In FIG. 14, feature amount attribute information is information indicating a position where the feature amount is estimated to have occurred.
The collation unit 1106 collates the feature amount specified by the feature amount specifying unit 1104 with the feature amount stored in the database unit 1107, and performs individual identification. The collation unit 1106 has a correlation value indicating a correlation between the feature quantity identified by the feature quantity identification unit 1104 and the feature quantity accumulated in the database unit 1107 above a threshold value or below a threshold value below a certain distance between the feature quantities. When it becomes, it determines with the similarity of those feature-values being high. In this case, the matching unit 1106 determines that the individuals associated with each feature amount are the same.
The collation unit 1106 calculates the correlation value or the distance between the feature amounts for all the feature amounts stored in the database unit 1107 for the feature amounts specified by the feature amount specifying unit 1104. Then, the collation unit 1106 determines whether or not the correlation value for all the feature amounts stored in the database unit 1107 is equal to or less than a threshold value or a distance between feature amounts is equal to or greater than a threshold value. When the correlation value for all the feature values stored in the database unit 1107 is equal to or less than a certain threshold value or the distance between feature values is equal to or greater than a certain threshold value, the matching unit 1106 performs the following processing. That is, the collation unit 1106 determines that the target individual is not registered in the database unit 1107. When the collation unit 1106 determines that the target individual is not registered, the database unit 1107 associates the feature quantity specified by the feature quantity specifying unit 1104 with information indicating the individual and the database addition processing unit 1105 described later. Register with
Thereby, when the individual exists in the real space after the next time, the collation unit 1106 can perform identification processing of the individual by performing collation using the feature amount registered in the database unit 1107.
The database addition processing unit 1105 receives information indicating an individual that the collation unit 1106 has determined to be unregistered in the database unit 1107 and a feature amount used to identify the individual from the collation unit 1106. The database addition processing unit 1105 also receives the attribute information of the received feature amount. Then, the database addition processing unit 1105 stores the received information in the database unit 1107 in association with each other.
FIG. 4 is a flowchart showing the operation of the individual identification system 1 in the first embodiment for carrying out the present invention.
The image information acquisition unit 1001 acquires image information included in environment information in the real space. The voice information acquisition unit 1002 acquires voice information included in the environment information in the real space (step S1). The sensor unit 1000 outputs the acquired environment information, that is, image information and audio information to the individual detection unit 1101.
The individual detection unit 1101 analyzes image information and audio information output from the sensor unit 1000, detects an individual to be identified, and specifies position information of the individual. The individual detection unit 1101 outputs information indicating the detected individual and the position information of the individual to the feature amount extraction unit 1102 and the feature amount determination unit 1103. Further, for each detected individual, the individual detection unit 1101 includes information indicating the individual, position information of the individual, image information of at least a part of the environment information corresponding to the position information, and at least one of the environment information. Are output to the feature quantity extraction unit 1102 (step S2). Specifically, the individual detection unit 1101 performs object detection processing based on a background difference method and pattern matching on image information, three-dimensional position detection processing using a multi-viewpoint image, and position detection of a sounding object by sound using a directional microphone. Use a combination of processing. Then, the individual detection unit 1101 detects an individual to be identified by the above-described combined process, and specifies position information of the individual.
The feature quantity extraction unit 1102 extracts a feature quantity for identifying the individual from the image information and audio information received from the individual detection unit 1101 and outputs the feature quantity to the feature quantity identification unit 1104 (step S3). Specifically, the feature amount extraction unit 1102 may extract a feature amount based on the color, shape, size, pattern, etc. of the object that can be acquired from the image information. Alternatively, the feature amount extraction unit 1102 may extract a feature amount based on an action of an object that can be acquired from video information. Alternatively, the feature amount extraction unit 1102 may extract a feature amount based on sound emitted by an object that can be acquired from audio information.
The feature quantity discriminating unit 1103 uses the image information and audio information corresponding to the individual input from the individual detection unit 1101 and the position information of the individual, and each of the feature quantities extracted by the feature quantity extracting unit 1102 is generated. Is discriminated (step S4). Specifically, the feature quantity discriminating unit 1103 considers the relation between the individual spatial position and the direction of the audio signal, the relation between the individual movement obtained from the video information and the movement direction of the audio signal, and the like. Identify the individual in which each occurred. Then, the feature amount determination unit 1103 outputs the determination result to the feature amount specifying unit 1104.
Further, the feature amount determination unit 1103 calculates an effective value representing the validity of the individual image information and audio information input from the individual detection unit 1101, and outputs the effective value to the feature amount specifying unit 1104 (step S5). As an effective value calculation method, for example, a method of calculating quantitatively using information such as signal SN ratio and signal strength, a method of calculating by a neural network or a support vector machine using previously collected learning data and so on.
The feature quantity specifying unit 1104 uses a plurality of types of feature quantities input from the feature quantity extraction unit 1102, the discrimination results input from the feature quantity discrimination unit 1103, and the effective value of each signal to identify the target individual. An effective feature amount is specified (step S6). The identified feature amount is output to the matching unit 1106. The collation unit 1106 collates the feature quantity input from the feature quantity identification unit 1104 with a plurality of types of feature quantities stored in the database unit 1107, and calculates a collation score indicating the correlation between the two feature quantities (step S7). ). The matching unit 1106 calculates a correlation value or a distance for each set of feature values, and calculates a matching score by integrating them. As a method of integrating correlation values or distances, there are a method of taking an average value of each value, a method of taking a maximum value, or addition or multiplication of each value. Further, as a method of integrating correlation values or distances, there is a method of integrating by a neural network or a support vector machine using learning data prepared in advance. When a system having this function is constructed, at least one of the integration methods may be implemented.
The collation unit 1106 determines whether or not the calculated collation score is equal to or less than a first threshold value (or greater than the first threshold value when the distance-based collation score is used). That is, the collation unit 1106 determines whether there is an individual corresponding to the database unit 1107 (step S8). When the collation unit 1106 determines that there is no individual corresponding to the database unit 1107 (in the case of “YES” (unregistered) in step S8), the database addition processing unit 1105 executes the following processing. That is, the database addition processing unit 1105 newly registers the feature amount specified by the feature amount specifying unit 1104 in the database unit 1107 (step S9). Thereby, individual identification after the next time is attained.
On the other hand, when the calculated matching score is equal to or higher than the first threshold (or lower than the first threshold when the distance-based matching score is used), the matching unit 1106 determines that there is an individual corresponding to the database unit 1107 ( If “NO” (registered) in step S8). In this case, the collation by the collation unit 1106 is completed.
The collation unit 1106 may have the following functions. First, the collation unit 1106 determines whether or not the calculated collation score is significantly higher than the first threshold value. When the calculated matching score is a distance-based matching score, the matching unit 1106 determines whether or not the calculated matching score is significantly smaller than the first threshold value. Here, the second threshold indicates a value larger than the first threshold (a value smaller than the first threshold when the matching score is a distance-based matching score).
For example, the collation unit 1106 determines whether or not the calculated collation score is greater than or equal to the second threshold value based on a predetermined second threshold value. The collation unit 1106 determines that the individual collation reliability is extremely high when the collation score is equal to or greater than a predetermined second threshold. Furthermore, the collation unit 1106 adds the individual feature amount to the database unit 1107 when the collation score is equal to or greater than a predetermined second threshold.
When the collation unit 1106 has the above function, the feature amount specifying unit 1104 may determine whether or not a feature amount associated with information indicating a certain individual is stored in the database unit 1107. When the feature amount specifying unit 1104 determines that the feature amount is stored in the database unit 1107, the feature amount having attribute information corresponding to the attribute information of the feature amount is used for the above-described individual identification processing. You may specify with quantity.
After the collation is completed by the collation unit 1106, the individual identification system 1 determines whether there is an individual that has not yet been identified among the individuals detected by the individual detection unit 1101 (step S10). When the individual identification system 1 determines that there is an individual that has not yet been identified (if “YES” in step S10 (there is an unidentified individual)), the individual steps from the feature determination process in step S4 are performed. repeat. By this operation, the individual identification system 1 can identify all the individuals existing in the real space.
On the other hand, when the individual identification system 1 determines that the identification process has been executed for all the individuals detected by the individual detection unit 1101 (in the case of “NO” (no unidentified individual) in step S10), the identification process is performed. finish.
The individual identification system 1 in the first embodiment determines an individual in which each of the feature amounts extracted from the environmental information has occurred. The individual identification system 1 obtains an effective value indicating the quality of the feature value for each feature value extracted from the environment information. Then, the individual identification system 1 specifies the feature amount used for the individual identification process for each individual detected from the environment information based on the determined result and the calculated effective value.
Therefore, the individual identification system 1 according to the first embodiment uses a feature value having a high effective value, that is, a higher quality feature amount among the feature amounts associated with each individual, and uses the feature amount in the individual identification process. Can do. As a result, the individual identification system 1 in the first embodiment can specify an appropriate feature amount to be used for individual identification for each of a plurality of identification targets. As a result, the individual identification system 1 can perform more accurate individual identification processing even when a plurality of individuals must be identified at the same time while being robust against environmental changes.
In the first embodiment, the feature amount determination unit 1103 may be included in the individual detection unit 1101. In this case, when detecting the individual to be identified from the image information and audio information received from the sensor unit 1000, the individual detection unit 1101 includes image information or audio information in a predetermined region including the position of the individual and the individual. May be associated with each other. Then, the individual detection unit 1101 calculates an effective value of the feature amount included in the image information or audio information based on the signal intensity of the image information or audio information associated with the individual. The individual detection unit 1101 may output information indicating the individual, position information of the individual, and image information and audio information associated with the individual to the feature amount extraction unit 1102. The feature amount extraction unit 1102 extracts feature amounts included in the received image information and audio information, respectively. Then, the feature quantity extraction unit 1102 associates the feature quantity, the attribute information of the feature quantity, and the effective value corresponding to the image information or audio information from which the feature quantity is extracted for each extracted feature quantity. You may output to 1104.
[Second Embodiment]
FIG. 5 is a block diagram showing a configuration of the individual identification system 2 according to the second embodiment of the present invention. As shown in FIG. 5, the individual identification system 2 according to the second embodiment of the present invention is different from the individual identification system 1 according to the first embodiment in the composite sensor unit 2200. Other components of the individual identification system 2 are the same as those of the individual identification system 1 in the first embodiment. The name of the audio / video acquisition unit 2000 is changed from that of the sensor unit 1000 in order to avoid confusion with the composite sensor unit 2200, but the configuration of the audio / video acquisition unit 2000 is the same as that of the sensor unit 1000 according to the first embodiment. The configuration is the same. The same components as those of the individual identification system 1 in the first embodiment of the individual identification system 2 are denoted by the same reference numerals as those in FIG. 1, and detailed description thereof is omitted.
In the second embodiment, the composite sensor unit 2200 includes a shape measuring unit 2201, a weight measuring unit 2202, a calorie measuring unit 2203, a speed measuring unit 2204, an optical property measuring unit 2205, and an odor measurement as shown in FIG. One or more of the part 2206 and the material inspection part 2207 are provided. The composite sensor unit 2200 is connected to the individual detection unit 1101. That is, the difference of the second embodiment from the first embodiment is that the composite sensor unit 2200 is added to the configuration of the individual identification system. Other components of the individual identification system 2 are the same as those of the individual identification system 1 in the first embodiment.
The shape measuring unit 2201 acquires information about the three-dimensional shape and volume of the individual. The weight measuring unit 2202 measures the weight of the individual. The calorie measurement unit 2203 measures the temperature of the individual. The speed measuring unit 2204 measures the speed of the individual. The optical characteristic measurement unit 2205 measures optical characteristics such as reflectance, transmittance, and refractive index of the solid surface. The odor measuring unit 2206 measures the odor of the individual. The material inspection unit 2207 acquires information such as hardness and material of the individual surface by infrared spectroscopy, ultrasonic inspection, or the like.
In addition, various sensor information acquired by the composite sensor unit 2200 is output to the feature amount extraction unit 1102 in the same manner as individual image information and audio information. A feature amount extraction unit 1102 extracts feature amounts from input image information, audio information, and various sensor information. The feature quantity discriminating unit 1103 performs processing for associating the feature quantity extracted from the sensor information with the individual and calculating the effective value of each feature quantity.
FIG. 7 is a flowchart showing an example of the operation of the individual identification system 2 in the second embodiment for carrying out the present invention. About the operation | movement similar to the identification system 1 in 1st Embodiment, the code | symbol same as FIG. 4 is attached | subjected and detailed description is abbreviate | omitted.
The second embodiment is different from the first embodiment in the following points. The first point is a point where image information and audio information included in the environment information as sensing information and sensing data acquired from the composite sensor unit 2200 are given as inputs to the individual detection unit 1101 (step S21). The second point is that when the individual detection unit 1101 performs individual detection from sensing information, sensor information obtained from the composite sensor is also used in addition to image information and audio information (step S22).
In the individual identification system 2 according to the second embodiment, the composite sensor unit 2200 acquires individual features that are difficult to acquire from normal image information and audio information. Such individual characteristics can be very useful information for individual identification. The individual identification system 2 according to the second embodiment can further improve the identification performance by using a combination of feature quantities other than the feature quantities obtained from image information and audio information.
[Third Embodiment]
FIG. 8 is a block diagram showing the configuration of the individual identification system 3 according to the third embodiment of the present invention. As shown in FIG. 8, the individual identification system 3 according to the third embodiment of the present invention is different from the individual identification system 1 according to the first embodiment in a biological information acquisition unit 3200 and a person detection unit 3101. That is, the individual detection unit 1101 of the identification processing device 1100 in the first embodiment is changed to the person detection unit 3101 of the identification processing device 3100 in the third embodiment. Other components of the individual identification system 3 are the same as those of the individual identification system 1 in the first embodiment. The same components as those of the individual identification system 1 in the first embodiment of the individual identification system 3 are denoted by the same reference numerals as those in FIG. 1, and detailed description thereof is omitted.
The biometric information acquisition unit 3200 of the individual identification system 3 according to the third embodiment includes the components shown in FIG. That is, the biometric information acquisition unit 3200 includes an iris pattern acquisition unit 3201, a fingerprint pattern acquisition unit 3202, a palmprint pattern acquisition unit 3203, a vein pattern acquisition unit 3204, a dentition pattern acquisition unit 3205, an auricle pattern acquisition unit 3206, and a gene sequence information acquisition. One or more of the units 3207 are provided. The biometric information acquisition unit 3200 is connected to the feature amount extraction unit 1102 and the feature amount determination unit 1103.
An iris pattern acquisition unit 3201 mainly includes an infrared light source and an infrared camera, and acquires an iris pattern of a person. When the iris pattern can be acquired by the video camera constituting the image information acquisition unit 1001, the image information acquisition unit 1001 may acquire the iris pattern.
A fingerprint pattern acquisition unit 3202 acquires a fingerprint pattern of a person. The fingerprint pattern acquisition unit 3202 may be configured to acquire a fingerprint pattern using a contact sensor, or may be configured to acquire a fingerprint pattern in a non-contact manner using a camera or the like.
A palm print pattern acquisition unit 3203 acquires a palm print and a palm pattern of a person. Similarly to the fingerprint pattern acquisition unit 3202, the palm pattern acquisition unit 3203 may be configured to acquire a palm pattern and a palm pattern by a contact sensor, or acquire a palm pattern and a palm pattern without contact with a camera or the like. Such a configuration may be adopted. Further, the palm print pattern acquisition unit 3203 may be configured to simultaneously acquire a fingerprint pattern.
The vein pattern acquisition unit 3204 acquires a vein pattern of a person. The vein pattern acquisition unit 3204 may be configured to acquire a vein pattern from a part such as a finger, palm, back of the hand, face, or neck, or may be configured to acquire a vein pattern from another part. Also good. Further, the fingerprint pattern acquisition unit 3202 or the palm print pattern acquisition unit 3203 may also acquire the vein pattern at the same time.
The dentition pattern acquisition unit 3205 acquires the shape and arrangement pattern of human teeth. The dentition pattern acquisition unit 3205 may acquire three-dimensional shape information as a dentition pattern in addition to image information captured by a camera.
An auricle pattern acquisition unit 3206 acquires a shape pattern of a human ear. Similar to the dentition pattern acquisition unit 3205, the auricle pattern acquisition unit 3206 may be configured to acquire three-dimensional shape information in addition to image information.
The gene sequence information acquisition unit 3207 acquires gene sequence information of a person. The gene sequence information acquisition unit 3207 may be configured to acquire gene sequence information from human skin, body hair, body fluid, or the like.
The person detection unit 3101 detects a person from the image information acquired by the sensor unit 1000. The person detection unit 3101 detects a person by using face detection for image information, walking pattern detection for video information, and the like. The person detection unit 3101 also has a function of detecting a person from image information and acquiring a person's face pattern and walking pattern.
In addition, the various sensor information acquired by the biological information acquisition unit 3200 is output to the feature amount extraction unit 1102 in the same manner as the image information and audio information acquired by the sensor unit 1000. Then, the feature quantity extraction unit 1102 performs feature quantity extraction from the various sensor information that is input. The feature quantity discriminating unit 1103 associates each feature quantity extracted from the sensor information with an individual and performs an effective value calculation process for each feature quantity.
FIG. 10 is a flowchart showing an example of the operation of the individual identification system 3 in the third embodiment for carrying out the present invention. About the operation | movement similar to 1st Embodiment, the code | symbol same as FIG. 4 is attached | subjected and detailed description is abbreviate | omitted.
The third embodiment differs from the first embodiment in the following points. The first point is that the identification target is limited to humans, so that the individual detection unit receives as input the image information acquired by the sensor unit 1000 as the sensing information, the voice information, and the biological information acquired by the biological information acquisition unit 3200. 1101 (step S31). The second point is a point in which the person detection unit 3101 performs a process of detecting a person from the sensing information, and acquires the face pattern and walking pattern of the person detected at the same time (step S32).
The individual identification system 3 according to the third embodiment performs individual identification using human biological information effective for personal identification by limiting the identification target from a general object to a human. The sensor unit 1000 acquires image information and audio information, and the biological information acquisition unit 3200 acquires biological information of a person. In addition, the person detection unit 3101 acquires a face pattern and a walking pattern as biometric information. Personal authentication technology using human biometric information has been widely studied as a related technology. Therefore, the related technique should just be utilized for each biometric authentication demonstrated by this Embodiment.
The individual identification system 3 in the third embodiment associates feature amounts extracted from image information, audio information, and other biological information with a person, and calculates an effective value of each feature amount. Then, the individual identification system 3 in the third embodiment uses the feature amount used for the individual identification processing of the person for each person detected by the person detection unit 3101 based on both the association and the effective value. Identify. The individual identification system 3 can prioritize a feature value having a high effective value, that is, a higher quality feature value among the feature amounts associated with each person, and can use the feature amount for personal identification processing.
As a result, the individual identification system 3 in the third embodiment can specify an appropriate feature amount for use in individual identification for each of a plurality of identification objects. As a result, the individual identification system 3 can perform more accurate personal identification processing even in the case where a plurality of persons must be identified at the same time while being robust against environmental changes.
[Fourth Embodiment]
FIG. 11 is a block diagram showing a configuration of the individual identification system 4 according to the fourth embodiment of the present invention.
As shown in FIG. 11, the individual identification system 4 in the fourth exemplary embodiment of the present invention includes a sensor unit 4000, an identification processing device 4100, and a database unit 4107. The sensor unit 4000 and the identification processing device 4100, and the identification processing device 4100 and the database unit 4107 are connected to be communicable with each other.
The sensor unit 4000 is a video camera that acquires video information and a microphone that acquires audio information, and each includes communication means such as a network interface. The sensor unit 4000 may be composed of a video camera that can simultaneously acquire video information and audio information.
The identification processing device 4100 is a computer, and includes at least communication means such as a CPU (Central Processing Unit), a memory, and a network interface. The identification processing device 4100 may include a reading device for reading a computer-readable recording medium such as a flexible disk or a CD-ROM (Compact Disc Read Only Memory), or a magnetic storage device. The identification processing device 4100 expands the program code received from the network interface on the memory, or reads the program code stored in the CD-ROM or the magnetic storage device and expands it on the memory. The identification processing device 4100 interprets and executes the developed program code so that the person detection unit 4101, the feature amount extraction unit 4102, the feature amount determination unit 4103, the feature amount identification unit 4104, the database, as shown in FIG. The additional processing unit 4105 and the collation unit 4106 operate as a computer that implements various functions. In the fourth embodiment, the identification processing device 4100 is a so-called personal computer (PC).
The database unit 4107 includes at least communication means such as a network interface and a magnetic storage device. The magnetic storage device stores facial feature amount information and voice feature amount information of a plurality of persons. The magnetic storage device stores at least one or more of face feature amount information and voice feature amount information per person. The magnetic storage device may store a plurality of both feature quantities per person. Further, the facial feature amount information and the voice feature amount information may be managed by a relational database management system (RDBMS). The fourth embodiment is an embodiment in which feature amounts are a face pattern and a voice pattern. And the individual identification system 4 in 4th Embodiment is naturally applicable also to the individual identification system using another feature-value.
In the fourth embodiment, there are two persons registered in the database unit 4107 (hereinafter referred to as person A and person B), and in the situation where only the person A speaks, the individual identification system 4 has the person A. And an example of an embodiment showing a case where person B is identified.
FIG. 12 is a flowchart showing the operation of the individual identification system 4 in the fourth embodiment for carrying out the present invention.
The sensor unit 4000 acquires the video information in which the person A and the person B are reflected and the voice information of the utterance, and passes them to the person detection unit 4101 (step S41).
The person detection unit 4101 detects that there are two persons in the space by processing such as face detection, and passes the detected face image data and audio data to the feature amount extraction unit 4102 and the feature amount determination unit 4103 ( Step S42).
The feature amount extraction unit 4102 extracts feature amounts for personal identification from the face image data and audio data acquired from the person detection unit 4101 (step S43).
The feature amount discriminating unit 4103 uses the face image data acquired from the person detecting unit 4101 to specify the speaker by speaker estimation or the like, and associates the speaker's face image and voice data with the speaker (step S44). Also, the feature amount discriminating unit 4103 calculates the effectiveness of the feature amount extracted from the face image or audio signal from the state of the face image and audio signal acquired from the person detection unit 4101 (step S45).
Based on the feature quantity acquired by the feature quantity extraction unit 4102 based on the correspondence between the feature quantity and the speaker obtained by the feature quantity discrimination unit 4103 and the effectiveness of each feature quantity, the feature quantity specifying unit 4104 The feature amount used for collation is identified and appropriately weighted (step S46).
The collation unit 4106 collates the feature quantity identified by the feature quantity identification unit 4104 with the feature quantity stored in the database unit 4107, and calculates a collation score of one person (step S47).
In the fourth embodiment, since it is assumed that two people are registered in the database, the database addition processing by the database addition processing unit 4105 is not executed (“NO” in step S48 (registered). )in the case of). Thereafter, since there is another unidentified person, the process for collating the other person (after step S44) is repeatedly executed ("YES" in step S50 (there is an unidentified person)). in the case of). When the collation is completed for both of the two persons, there is no unidentified person, and the process ends (in the case of “NO” (no unidentified person) in step S50).
In the personal recognition system described in Patent Literature 1, image information and audio information acquired by an acquisition unit that acquires environmental information of a space where a person exists is limited to a single object. Therefore, in the personal recognition system described in Patent Document 1, it is difficult to specify an appropriate feature amount for accurate identification in the situation assumed in the fourth embodiment. On the other hand, the individual identification system 4 according to the fourth embodiment of the present invention is suitable for use in individual identification for each of a plurality of identification targets even in a situation where there are two or more persons. A feature amount can be specified.
[Fifth Embodiment]
FIG. 13 is a block diagram showing a configuration of the feature quantity specifying device 50 according to the fifth embodiment of the present invention.
Referring to FIG. 13, the feature quantity specifying device 50 according to the present embodiment includes a registration unit 5001, an effective value calculation unit 5002, and a feature quantity specifying unit 5003.
The registration unit 5001 includes information indicating an individual detected from environment information that is information resulting from the environment of the space where the individual to be subjected to the individual identification process exists, and position information indicating the position of each individual. receive. Further, the registration unit 5001 receives the feature amount extracted from the environment information and attribute information indicating each attribute of the feature amount. Then, the registration unit 5001 determines an individual in which each of the received feature amounts has occurred based on the received position information and attribute information.
The effective value calculation unit 5002 obtains an effective value indicating the quality of the feature value for each of the feature values received by the registration unit 5001.
The feature amount specifying unit 5003 performs the following processing based on the determination result by the registration unit 5001 and the effective value calculated by the effective value calculating unit 5002. That is, the feature quantity specifying unit 5003 specifies the feature quantity used for the individual identification process for each individual specified by the information indicating the individual received by the registration unit 5001.
The feature quantity specifying device 50 according to the present embodiment is based on attribute information indicating the type of an individual to be subjected to individual identification processing and attribute information indicating the type of feature quantity extracted from environment information. Associate quantity with. Therefore, the feature quantity specifying device 50 of this embodiment can specify an appropriate feature quantity to be used for individual identification for each of a plurality of identification targets.
As mentioned above, each embodiment described so far is a preferred embodiment of the present invention, and the scope of the present invention is not limited only to the above-described embodiment. Each embodiment can be implemented in various forms without departing from the gist of the present invention.
For example, as an individual identification system, an embodiment in which the first embodiment, the second embodiment, the third embodiment, the fourth embodiment, and the fifth embodiment are combined with each other is configured. May be.
In addition, each component in each embodiment of the present invention can be realized by a computer and a program as well as its function in hardware. The program is provided by being recorded on a computer-readable recording medium such as a magnetic disk or a semiconductor memory, and is read by the computer when the computer is started up. The read program causes the computer to function as a component in each of the embodiments described above by controlling the operation of the computer. In this case, the computer includes a central processing unit (CPU) on which the individual identification program is read and executed, a storage device (such as a hard disk) that stores feature quantities as a database, and input means such as a camera and a microphone.
Taking the first embodiment of the present invention as an example, the individual identification program read into the CPU causes the computer to function as the identification processing device 1100 described in the first embodiment.
An example of the effect of the present invention is that it is possible to specify an appropriate feature amount for use in individual identification for each of a plurality of identification targets.
This application claims the priority on the basis of Japanese application Japanese Patent Application No. 2010-178644 for which it applied on August 9, 2010, and takes in those the indications of all here.
A part or all of the above embodiment can be described as in the following supplementary notes, but is not limited thereto.
(Appendix 1)
Extracted from the environmental information, information indicating the individual detected from the environmental information, which is information resulting from the environment of the space where the individual subject to the individual identification process exists, position information indicating the position of the individual, and the environmental information And the attribute information indicating each attribute of the feature quantity is received, and the individual in which the feature quantity has occurred is determined for each feature quantity based on the position information and the attribute information. A registration section;
For each feature quantity extracted from the environment information, an effective value calculation unit for obtaining an effective value indicating the quality of the feature quantity;
Based on the determination result and the effective value, for each detected individual, a feature amount specifying unit that specifies a feature amount used for individual identification processing of the individual;
A feature quantity specifying device.
(Appendix 2)
For each individual, the feature amount specifying unit assigns a feature amount having an effective value equal to or greater than a predetermined threshold among the feature amounts determined to be generated from one individual based on the determination result to the one individual. The feature amount specifying apparatus according to claim 1, wherein the feature amount is specified as a feature amount used for individual identification processing.
(Appendix 3)
For each feature amount extracted from the environment information, the registration unit calculates, for each individual, a probability that the feature amount has occurred from each individual,
The feature amount according to claim 1 or 2, wherein the feature amount specifying unit specifies, for each detected individual, a feature amount used for individual identification processing of the individual based on the effective value and the probability. Specific device.
(Appendix 4)
The feature quantity specifying unit according to claim 3, wherein the feature quantity specifying unit specifies a feature quantity having a product of the effective value and the probability equal to or greater than a predetermined threshold as a feature quantity used for individual identification processing of the one individual. apparatus.
(Appendix 5)
The attribute information of the feature amount includes position information indicating a position where the feature amount has occurred,
The registration unit includes:
Based on the difference between the position specified based on the position information of each individual and the position specified based on the attribute information of each feature amount, each of the feature amounts is a feature amount generated from the individual. The feature quantity specifying device according to claim 3, wherein certain probabilities are respectively calculated.
(Appendix 6)
The registration unit includes:
The information indicating the time when the individual is detected is received together with the position information of the individual, the position information is associated with the information indicating the time, and the feature amount is specified together with the attribute information of the feature amount The attribute information is associated with the information indicating the time, and the first individual is based on the position information of the first individual and the information indicating the time associated with each position information. A vector indicating the time change of the position of the first feature value is generated, and the time change of the position of the first feature value is calculated based on the attribute information of the first feature value and the information indicating the time associated with each attribute information. The vector shown is generated, and the length of the vector specified based on the difference between the vector generated based on the second individual and the vector generated based on the second feature amount is less than a predetermined value In case, Feature quantity of the second discriminates the feature quantity generated from the second individual, the feature identification apparatus according to any one of claims 1 to 5.
(Appendix 7)
The feature quantity specifying device includes a database unit that stores a matching feature quantity used for individual identification, attribute information of the feature quantity, and information indicating the individual in association with each other.
When the feature amount associated with information indicating one individual is stored in the database unit, the feature amount specifying unit, when the feature amount has attribute information corresponding to the attribute information of the feature amount, The feature amount specifying apparatus according to claim 1, wherein the feature amount specifying device specifies the feature amount used for the individual identification processing of the one individual.
(Appendix 8)
An environment information acquisition unit for acquiring environment information, which is information resulting from the environment of a space where a plurality of individuals to be subjected to individual identification processing exist;
An individual detection unit for detecting the plurality of individuals from the environment information together with position information indicating the position of each individual;
A feature amount extraction unit that extracts a plurality of feature amounts from the environment information together with attribute information indicating the type of each feature amount;
A registration unit that determines, for each feature quantity, an individual in which the feature quantity has occurred, based on the position information of each of the individuals and the attribute information of each of the feature quantities;
For each feature quantity extracted from the environment information, an effective value calculation unit for obtaining an effective value indicating the quality of the feature quantity;
Based on the determination result and the effective value, for each detected individual, a feature amount specifying unit that specifies a feature amount used for individual identification processing of the individual;
A collation unit that performs individual identification processing for each individual extracted based on the identified feature amount;
An individual identification system comprising:
(Appendix 9)
Extracted from the environmental information, information indicating the individual detected from the environmental information, which is information resulting from the environment of the space where the individual subject to the individual identification process exists, position information indicating the position of the individual, and the environmental information And the attribute information indicating the attribute of each of the feature quantities is received, and the individual in which the feature quantity has occurred is determined for each feature quantity based on the position information and the attribute information. ,
For each feature quantity extracted from the environment information, an effective value indicating the quality of the feature quantity is obtained,
A feature amount specifying method for specifying, for each detected individual, a feature amount used for individual identification processing of the individual based on the result of the determination and the effective value.
(Appendix 10)
Extracted from the environmental information, information indicating the individual detected from the environmental information, which is information resulting from the environment of the space where the individual subject to the individual identification process exists, position information indicating the position of the individual, and the environmental information And the attribute information indicating each attribute of the feature quantity is received, and the individual in which the feature quantity has occurred is determined for each feature quantity based on the position information and the attribute information. Processing,
For each feature quantity extracted from the environment information, a process for obtaining an effective value indicating the quality of the feature quantity;
A program for causing a computer to execute, for each detected individual, a feature amount used for individual identification processing of the individual based on the determination result and the effective value.
(Appendix 11)
The attribute information of the feature amount includes position information indicating a position where the feature amount has occurred,
The registration unit includes:
When the difference between the position specified based on the position information of one individual and the position where the one feature amount is less than or equal to a predetermined threshold, the one feature amount is a feature amount generated from the one individual. The feature quantity specifying device according to supplementary note 1, wherein the feature quantity specifying device is specified.
(Appendix 12)
The feature quantity specifying unit includes:
Storing associating information, which is information associating individual position information with feature amount attribute information;
The feature quantity specification according to appendix 1, wherein the feature quantity having the attribute information of the feature quantity corresponding to the association information including the position information of one individual is identified as a feature quantity used for the individual identification process of the one individual. apparatus.
(Appendix 13)
The individual position information includes information indicating the movement direction of the individual,
The attribute information of the feature amount includes information indicating whether the volume is increasing or decreasing,
The feature quantity specifying unit includes:
When the attribute information of the first individual includes information indicating the first direction, the feature amount having the attribute information including the information whose volume is increased is used for the individual identification processing of the first individual And identify
When the attribute information of the second individual includes information indicating a second direction that is the opposite direction to the first direction, the feature amount having attribute information including information in which the volume is decreased The feature amount specifying device according to attachment 1, wherein the feature amount is specified as a feature amount used for individual identification processing of the individual.
(Appendix 14)
An environmental information acquisition unit for acquiring the environmental information;
Detecting the individual from the environmental information together with attribute information indicating the type of each individual, and outputting them to the registration unit;
Extracting the feature value from the environment information together with attribute information indicating the type of each feature value, and outputting the feature value to the registration unit;
The feature quantity identifying device according to appendix 1, comprising:
(Appendix 15)
The individual identification system includes a database unit that stores a matching feature amount used for individual identification, attribute information of the feature amount, and information indicating the individual in association with each other.
The collation unit reads a feature quantity for collation used for identifying an individual associated with the feature quantity based on the feature quantity identified by the feature quantity identification unit, and identifies the feature quantity. A collation score indicating a correlation between the feature quantity identified by the unit and the feature quantity read from the database unit, and when the collation score is equal to or greater than a predetermined threshold, the feature quantity identified by the feature quantity identification unit The information indicating the individual and the attribute information of the feature amount are associated with each other and stored in the database unit,
When the feature amount associated with information indicating one individual is stored in the database unit, the feature amount specifying unit, when the feature amount has attribute information corresponding to the attribute information of the feature amount, The individual identification system according to appendix 8, wherein the individual identification system identifies the feature amount used for the individual identification processing of the one individual.

DESCRIPTION OF SYMBOLS 1 Individual identification system 2 Individual identification system 3 Individual identification system 4 Individual identification system 50 Feature-value identification apparatus 1000 Sensor part 1001 Image information acquisition part 1002 Audio | voice information acquisition part 2000 Image audio | voice acquisition part 4000 Sensor part 1100 Identification processing apparatus 3100 Identification processing apparatus 4100 Identification processing device 1101 Individual detection unit 3101 Person detection unit 4101 Person detection unit 1102 Feature amount extraction unit 4102 Feature amount extraction unit 1103 Feature amount determination unit 4103 Feature amount determination unit 1113 Registration unit 5001 Registration unit 1123 Effective value calculation unit 5002 Effective value calculation unit 1104 Feature amount specifying unit 4104 Feature amount specifying unit 5003 Feature amount specifying unit 1105 Database addition processing unit 4105 Database addition processing unit 1106 Joint unit 4106 Verification unit 1107 Database unit 4107 Database unit 2200 Composite sensor unit 2201 Shape measurement unit 2202 Weight measurement unit 2203 Calorie measurement unit 2204 Speed measurement unit 2205 Optical property measurement unit 2206 Odor measurement unit 2207 Material inspection unit 3200 Biometric information acquisition unit 3201 Iris pattern acquisition unit 3202 Fingerprint pattern acquisition unit 3203 Palmprint pattern acquisition unit 3204 Vein pattern acquisition unit 3205 Dental pattern acquisition unit 3206 Auricular pattern acquisition unit 3207 Gene sequence information acquisition unit

Claims

Extracted from the environmental information, information indicating the individual detected from the environmental information, which is information resulting from the environment of the space where the individual subject to the individual identification process exists, position information indicating the position of the individual, and the environmental information And the attribute information indicating each attribute of the feature quantity is received, and the individual in which the feature quantity has occurred is determined for each feature quantity based on the position information and the attribute information. A registration section;
For each feature quantity extracted from the environment information, an effective value calculation unit for obtaining an effective value indicating the quality of the feature quantity;
Based on the determination result and the effective value, for each detected individual, a feature amount specifying unit that specifies a feature amount used for individual identification processing of the individual;
A feature quantity specifying device.
For each individual, the feature amount specifying unit assigns a feature amount whose effective value is equal to or greater than a predetermined threshold among the feature amounts determined to be generated from one individual based on the determination result to the one individual. The feature amount specifying device according to claim 1, wherein the feature amount is specified as a feature amount used for individual identification processing.
For each feature amount extracted from the environment information, the registration unit calculates, for each individual, a probability that the feature amount has occurred from each individual,
The feature amount according to claim 1 or 2, wherein the feature amount specifying unit specifies, for each detected individual, a feature amount used for individual identification processing of the individual based on the effective value and the probability. Specific device.
The feature quantity specifying unit according to claim 3, wherein the feature quantity specifying unit specifies a feature quantity having a product of the effective value and the probability equal to or greater than a predetermined threshold as a feature quantity used for individual identification processing of the one individual. apparatus.
The attribute information of the feature amount includes position information indicating a position where the feature amount has occurred,
The registration unit includes:
Based on the difference between the position specified based on the position information of each individual and the position specified based on the attribute information of each feature amount, each of the feature amounts is a feature amount generated from the individual. The feature quantity specifying device according to claim 3, wherein certain probabilities are respectively calculated.
The registration unit includes:
The information indicating the time when the individual is detected is received together with the position information of the individual, the position information is associated with the information indicating the time, and the feature amount is specified together with the attribute information of the feature amount The attribute information is associated with the information indicating the time, and the first individual is based on the position information of the first individual and the information indicating the time associated with each position information. A vector indicating the time change of the position of the first feature value is generated, and the time change of the position of the first feature value is calculated based on the attribute information of the first feature value and the information indicating the time associated with each attribute information. The vector shown is generated, and the length of the vector specified based on the difference between the vector generated based on the second individual and the vector generated based on the second feature amount is less than a predetermined value In case, Feature quantity of the second discriminates the feature quantity generated from the second individual, the feature identification apparatus according to any one of claims 1 to 5.
The feature quantity specifying device includes a database unit that stores a matching feature quantity used for individual identification, attribute information of the feature quantity, and information indicating the individual in association with each other.
When the feature amount associated with information indicating one individual is stored in the database unit, the feature amount specifying unit, when the feature amount has attribute information corresponding to the attribute information of the feature amount, The feature amount specifying apparatus according to claim 1, wherein the feature amount specifying device specifies the feature amount used for the individual identification processing of the one individual.
An environment information acquisition unit for acquiring environment information, which is information resulting from the environment of a space where a plurality of individuals to be subjected to individual identification processing exist;
An individual detection unit for detecting the plurality of individuals from the environment information together with position information indicating the position of each individual;
A feature amount extraction unit that extracts a plurality of feature amounts from the environment information together with attribute information indicating the type of each feature amount;
A registration unit that determines, for each feature quantity, an individual in which the feature quantity has occurred, based on the position information of each of the individuals and the attribute information of each of the feature quantities;
For each feature quantity extracted from the environment information, an effective value calculation unit for obtaining an effective value indicating the quality of the feature quantity;
Based on the determination result and the effective value, for each detected individual, a feature amount specifying unit that specifies a feature amount used for individual identification processing of the individual;
A collation unit that performs individual identification processing for each individual extracted based on the identified feature amount;
An individual identification system comprising:
Extracted from the environmental information, information indicating the individual detected from the environmental information, which is information resulting from the environment of the space where the individual subject to the individual identification process exists, position information indicating the position of the individual, and the environmental information And the attribute information indicating the attribute of each of the feature quantities is received, and the individual in which the feature quantity has occurred is determined for each feature quantity based on the position information and the attribute information. ,
For each feature quantity extracted from the environment information, an effective value indicating the quality of the feature quantity is obtained,
A feature amount specifying method for specifying, for each detected individual, a feature amount used for individual identification processing of the individual based on the result of the determination and the effective value.
Extracted from the environmental information, information indicating the individual detected from the environmental information, which is information resulting from the environment of the space where the individual subject to the individual identification process exists, position information indicating the position of the individual, and the environmental information And the attribute information indicating each attribute of the feature quantity is received, and the individual in which the feature quantity has occurred is determined for each feature quantity based on the position information and the attribute information. Processing,
For each feature quantity extracted from the environment information, a process for obtaining an effective value indicating the quality of the feature quantity;
A recording medium recording a program for causing a computer to execute, for each detected individual, a feature amount used for individual identification processing based on the determination result and the effective value. .