JP2012212071A - Face image authentication device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a face image authentication device which outputs a voice at an appropriate timing for each user when the faces of a plurality of users are verified sequentially.SOLUTION: A face image authentication device 1 comprises: a voice output part 3 for outputting a voice; face detection means 11 for extracting an input face image from an input image; face verification means 13 for verifying a registered face image with the input face image to determine whether they are the same person; voice output determination means 14 for determining whether a voice output part is outputting a voice; output voice-use information formation means 15 for forming output voice-use information; voice synthesis means 17 for synthesizing an output voice signal from the output voice information; and control means 16 for outputting a voice based on the output voice signal from the voice output part. The output voice-use information formation means forms, as the output voice information, standard voice-use information when the voice output part is outputting no voice, and forms abbreviated voice-use information shorter than the standard voice-use information when the voice output part is outputting a voice, both at a time when the face verification means performs the determination.

Description

  The present invention relates to a face image authentication apparatus, and more particularly to a face image authentication apparatus that outputs a sound to a user.

Conventionally, a user can enter a room by checking the face image of the face of a user (employee, etc.) that is installed at the entrance of a company's office and photographed with the face image registered in advance. Face image authentication devices that authenticate whether or not a person is a person have been developed. The conventional face image authentication device collates face images when the user temporarily stops in front of the entrance door of the living room and inputs a password. However, in recent years, a walking-type face image authentication device has been developed that captures and collates the face of a user who is walking and authenticates before the user approaches the entrance. In such a walk-type facial image authentication device, when a plurality of users approach the entrance of a living room one after another, such as when going to work, each user must authenticate before reaching the entrance. The user cannot enter the room smoothly, and convenience is impaired.
Therefore, Patent Document 1 proposes a person recognition device that changes the accuracy of face matching processing in accordance with the number of people passing to shorten the time of face matching processing. This person recognition device calculates the number of faces in the input image, and changes the resolution of the face image in the face matching process, the matching target area, etc. according to the number of faces, so that the number of people passing by increases. Control is performed so as to shorten the time of the hit face matching process.

JP 2007-156541 A

  The person recognition device described in Patent Document 1 can reduce the time for face matching processing per person when there are a large number of passing people, and each user can enter the room smoothly. However, in the walking type face image authentication device, for example, when trying to output a sound corresponding to the result of the face matching process to the user, the result of the face matching process is generally compared with the time required for the face matching process. The time for outputting the corresponding sound becomes very long. Therefore, when multiple users pass at the same time, even if the time required for the face matching process of each user is shortened, the influence on the time until the voice output for each user is completed is small. There is a possibility that the audio output may not be in time before entering the room.

  Accordingly, an object of the present invention is to provide a face image authentication apparatus capable of sequentially collating a plurality of users and optimizing the sound output timing for each user when outputting the sound to each user. It is to provide.

  The present invention for solving this problem includes an imaging unit that sequentially acquires input images obtained by photographing a person to be collated, a storage unit that stores a registered face image of the registrant in advance, and a voice that outputs sound to the person to be collated. Whether the output unit, the face detection means for extracting the face area image of the person to be collated as an input face image each time an input image is acquired, the registered face image and the input face image are collated, and whether or not they are the same person A face collating unit for determining the output, a sound output determining unit for determining whether or not the sound output unit is outputting sound, and an output sound for generating output sound information based on a determination result of the sound output determining unit Provided is a face image authentication device having information creation means, voice synthesis means for synthesizing an output voice signal from information for output voice, and control means for outputting the output voice signal from a voice output unit. In the face image authentication apparatus, the output voice information creating unit creates standard voice information as the output voice information when the voice output unit is not outputting the voice at the time of the determination by the face matching unit. If the voice output unit is outputting voice at the time of, short voice information is created that is shorter than the standard voice information.

  In the face image authentication apparatus according to the present invention, the storage unit further stores an authentication boilerplate and a personal name of the registrant in association with the registered face image of the registrant. The voice information is created from the authentication boilerplate and the personal name of the registrant who is determined to be the same person as the person to be collated in the input face image by the face matching means, and the shortened voice information is It is preferable to create it from the personal name of the registrant who has been determined by the verification means to be the same person as the verification target person shown in the input face image.

  In the face image authentication apparatus according to the present invention, the control means stores the output sound information in the storage unit every time the output sound information creation means creates the output sound information. When the audio output of the output audio signal is completed, the output audio information corresponding to the output audio signal is deleted from the storage unit, and the audio output determination means It is preferable that it is determined that the output unit is outputting audio, and if the output audio information is not stored in the storage unit, it is determined that the audio output unit is not outputting audio.

  Further, in the face image authentication device according to the present invention, the storage unit stores an attribute for each registrant, and a control unit, which is predetermined according to the attribute, outputs an output audio signal to the audio output unit. The priority that defines the order of output is stored, and the control means outputs the output audio signal for the output audio information for which the audio output by the audio output unit has not started among the output audio information stored in the storage unit. It is preferable to have the audio output unit output audio in descending order of priority.

  Since the face image authentication apparatus according to the present invention can sequentially collate a plurality of users and can optimize the voice output timing for each user when outputting the voice to each user, A person can recognize that he / she has been authenticated by listening to the sound output directed toward him / her, thereby improving the usability.

It is a schematic block diagram of the face image authentication apparatus to which this invention is applied. It is a schematic diagram showing the installation example of an imaging part in case a face image authentication apparatus is installed in the entrance of an office building. It is a schematic diagram of a registration table. (A) is a schematic diagram which shows the relationship between an input image and output audio | voice when a some user passes discretely to the channel | path in which the face image authentication apparatus was installed, (b) is a face image authentication apparatus It is a schematic diagram which shows the relationship between an input image and an output audio | voice when a some user passes along the channel | path in which was installed. It is a schematic diagram of an output table. It is a schematic diagram of a fixed phrase table. It is a flowchart which shows operation | movement of the registration process of the notification information by a face image authentication apparatus. It is a flowchart which shows the operation | movement of the audio | voice output process by a face image authentication apparatus. It is a schematic diagram which shows the relationship between an input image and output sound in case a some user passes along the channel | path in which the face image authentication apparatus of this embodiment was installed continuously. It is a schematic diagram of the output table in the modification of this invention. (A), (b) is a schematic diagram of the output table for demonstrating the order change of the audio | voice output according to the priority of a registrant.

Hereinafter, a face image authentication apparatus according to an embodiment of the present invention will be described with reference to the drawings.
The face image authentication apparatus to which the present invention is applied sequentially compares a plurality of verification target persons, and outputs a sound corresponding to the verification result to each verification target person. Then, when the face image authentication apparatus is already outputting the voice when the collation of the person to be collated is completed, the face image authentication apparatus shortens the newly output voice to be shorter than the normal output voice. Thereby, the face image authentication device aims to optimize the output timing of the sound for a plurality of verification subjects.

  FIG. 1 is a diagram showing a schematic configuration of a face image authentication apparatus 1 to which the present invention is applied. As illustrated in FIG. 1, the face image authentication device 1 includes an imaging unit 2, an audio output unit 3, an interface unit 4, a storage unit 5, and a processing unit 6. Hereinafter, each part of the face image authentication device 1 will be described in detail.

The imaging unit 2 is installed so that the face of the person to be collated can be photographed according to the environment in which the face image authentication device 1 is operated. And the imaging part 2 produces | generates the image in which the collation subject was shown as an input image. For this purpose, the imaging unit 2 includes a camera including an optical system that forms an image of the face of the person to be collated on a two-dimensional array of solid-state imaging elements such as a CCD or a CMOS.
The imaging unit 2 outputs the generated input image to the interface unit 4.

  The imaging unit 2 may create a color multi-tone image as an input image, or may be a camera having sensitivity in the near infrared region and creating a gray image. Moreover, it is preferable that the image pick-up element array which the image pick-up part 2 has has the pixel number of the grade which can distinguish facial features, such as eyes, a nose, and a mouth of a collation target person who are reflected in the input image.

  FIG. 2 schematically shows an installation example of the imaging unit 2 when the face image authentication device 1 is installed at the entrance of an office building. As shown in FIG. 2, for example, when the face image authentication device 1 is installed at the entrance 200 of an office building, the imaging unit 2 includes a wall on which the entrance 200 is installed so that the monitoring area includes a passage leading to the entrance 200. It is attached to the top or ceiling of the camera with the shooting direction slightly downward and toward the passage. Thereby, the imaging unit 2 can capture the verification target persons 202 to 204 heading toward the entrance 200 (heading in the traveling direction 201) at predetermined time intervals (for example, 200 msec).

  The audio output unit 3 is, for example, a speaker, is connected to the interface unit 4, emits audio corresponding to the audio signal received from the interface unit 4, and notifies the verification target person. As shown in FIG. 2, the audio output unit 3 is installed in the vicinity of the imaging unit 2 above or on the ceiling where the entrance 200 is installed, and the audio output direction is directed slightly downward and toward the passage side. It is attached in the state. Further, the audio output unit 3 is adjusted so as to output the sound at an appropriate volume so that the sound to be output can be heard by the person to be collated.

  The interface unit 4 is an interface circuit connected to the imaging unit 2 and the audio output unit 3, and includes an interface circuit conforming to a serial bus such as a video interface, an audio interface, or a universal serial bus. The interface unit 4 passes the input image received from the imaging unit 2 to the processing unit 6, and passes the audio signal received from the processing unit 6 to the audio output unit 3.

  The storage unit 5 includes at least one of a semiconductor memory, a magnetic recording medium and its access device, and an optical recording medium and its access device. The storage unit 5 stores a computer program, various parameters, data, and the like for controlling the face image authentication device 1. In addition, the storage unit 5 stores a registration table for storing registration information for each registrant, a fixed phrase table for storing information on the fixed phrases of speech output by the face image authentication apparatus 1, and the face image authentication apparatus 1 to output. An output table for managing information related to audio is stored. In addition, the storage unit 5 stores face data that is data related to the registrant's face image in the registration table. Details of these registration table, fixed phrase table, and output table will be described later.

The processing unit 6 includes one or a plurality of processors and their peripheral circuits. And the process part 6 controls the output of the audio | voice according to the collation result while performing the face collation process with respect to a collation subject. For this purpose, the processing unit 6 includes a face detection unit 11, a face tracking unit 12, a face matching unit 13, a voice output determination unit 14, and an output voice as functional modules implemented by software operating on the processor. Use information creation means 15, control means 16, and speech synthesis means 17.
Note that these units included in the processing unit 6 may be configured by independent integrated circuits, firmware, a microprocessor, and the like.
Hereinafter, each part of the processing unit 6 will be described in detail.

Each time the input image is acquired, the face detection unit 11 detects an input face area that is an area in which the face of the person to be collated is captured from the input image, cuts out the area from the input image, and extracts the input face image. create. For this purpose, the face detection unit 11 detects edge pixels from the input image using an edge detection filter such as a sobel filter. Then, the face detection unit 21 detects an elliptical outline that seems to be the outline of a human face by performing generalized Hough transform on the detected edge pixels, and sets an area surrounded by the outline as an input face area.
Alternatively, the face detection unit 11 may detect the input face area using an Adaboost classifier. This method is described in a paper by P. Viola and M. Jones “Rapid Object Detection Using a Boosted Cascade of Simple Features” (Proc. The IEEE International Conference on Computer Vision and Pattern Recognition, vol.1, pp.511-518, 2001). ) Can be referred to. In that case, the Adaboost discriminator uses a plurality of sample images in which a human face is reflected and a plurality of sample images in which a human face is not captured, and the face is reflected in an image in which the face is reflected. And an identification result indicating that a face is not captured is output in advance for an image in which a face is not captured. Then, the face detection means 11 cuts out an area of a predetermined size in the input image while shifting its position, inputs it to the Adaboost classifier, and obtains an identification result as to whether or not a face is reflected in the area. , Detect the input face area.
When a plurality of verification target persons are shown in the input image, the face detection unit 11 detects an input face area for each verification target person and creates an input face image.
When the face detection unit 11 successfully detects the input face area on the input image, the face detection unit 11 notifies the face tracking unit 12 of information representing the input face area.

The face tracking unit 12 performs a tracking process using a known tracking technique on the input face area detected from the face detection unit 11 over a plurality of input images continuously acquired at predetermined time intervals, and the same. Associate an input face area with a human face.
For example, the face tracking unit 12 detects the position of the center of gravity of the input face area (hereinafter referred to as the input face area of the current frame) detected from the latest input image and the input face area ( Hereinafter, the distance of the center of gravity position of the previous frame (referred to as the input face area of the previous frame) is obtained, and when the distance is equal to or less than a predetermined threshold, the input face area is associated with the same person. It should be noted that when the person to be collated moves a certain distance when moving away from the imaging unit 2 and when moving the same distance when moving closer to the imaging unit 2, input in the input image is performed before and after the movement. The difference in the position of the face area is different. For this reason, for example, by setting a predetermined threshold value as the size of the input face area, a person shown in each of the input face area in the current frame and the input face area in the previous frame regardless of the position of the person to be verified in the monitoring area Can be appropriately evaluated whether or not they are the same person. In addition, when a plurality of input face areas are extracted, the face tracking unit 12 checks whether or not the input face areas with the closest distance between the center of gravity positions correspond to each other.
Alternatively, the face tracking unit 12 may perform input face region tracking processing using a method such as an optical flow or a particle filter.

  The input face area that could not be associated with the input face area of the previous frame is set as an input face area corresponding to a new person to be collated, and is a target of subsequent tracking processing. Also, the input face area of the previous frame that could not be associated with all the input face areas of the current frame is excluded from the subsequent tracking processing targets.

  If the input face image cut out from the input face area that is the target of the tracking process by the face tracking means 12 has not been verified yet, the face matching means 13 is stored in the storage unit 5. Each registered face image in the registered table is collated to determine whether or not the image is from the same person.

FIG. 3 is a schematic diagram of a registration table that stores registration information for each registrant stored in the storage unit 5. In the registration table 300 shown in FIG. 3, one row corresponds to one registrant.
In each column of the leftmost column, identification information 301 of the registrant is shown. The identification information 301 is, for example, a registrant's user name, user identification number, or password. Alternatively, the identification information 301 may be anything as long as it can uniquely identify the registrant, such as an employee number or a continuous positive integer. Further, face data 302 is stored in each column of the second column from the left. The face data 302 is data related to the registrant's face image, and in this embodiment, a registered face image is stored as the face data 302.

  The face matching unit 13 can use various known matching methods as the matching process. For example, the face matching unit 13 performs pattern matching between the input face image extracted by the face extracting unit 11 and the registered face image. The face collating unit 13 calculates the sum of squares of the luminance values of the pixels of the input face image and the corresponding pixels of the registered face image while shifting the positions of the input face image and the registered face image, and most of the calculated sums of squares. The reciprocal of the normalized value obtained by dividing the smaller one by the number of pixels included in the input face image is obtained as the similarity. Then, when the highest similarity among the similarities obtained for each registered face image exceeds a predetermined collation threshold, the face collating unit 13 selects the person to be collated in the input face image with the highest similarity. It is determined that the registered person is a registered face image having a value (success in collation). On the other hand, the face collating means 13 determines that the person to be collated in the input face image is not a registrant (collation failure) when any similarity does not exceed a predetermined collation threshold. This collation threshold value is appropriately determined according to the environment, purpose, etc. in which the face image authentication device 1 is installed.

Alternatively, the face collating unit 13 may obtain the similarity based on face feature points that are characteristic parts of the face. In that case, the face collating means 13 extracts a plurality of facial feature points such as both eye corners, both eye region centers, nose tip, mouth point, mouth corner point from the input face image and the registered face image. For example, the face matching unit 13 performs edge extraction processing on the input face image and the registered face image, and extracts edge pixels having a large luminance difference from surrounding pixels. Then, the face collating unit 13 determines whether the feature amount obtained based on the position and pattern of the edge pixel satisfies a predetermined condition for parts such as both eye corners, both eye region centers, nose apex, mouth point, and mouth corner point. It is extracted as a face feature point by checking whether or not and specifying the position of each part. Then, the face collating unit 13 outputs, for each extracted face feature point, position information of each face feature point on the input face image and the registered face image (for example, two-dimensional with the upper left end of the input face image and the registered face image as the origin. (Coordinate value) is calculated. Then, the face collating unit 13 obtains the sum of the distances between corresponding feature points of the input face image and the registered face image as a positional deviation amount, and uses the reciprocal of the positional deviation amount as the similarity.
Alternatively, the face matching unit 13 calculates an average value of luminance or color difference for each extracted facial feature point for a local region near the facial feature point. In that case, the face matching unit 13 may obtain an absolute value of the difference between the calculated average values for each corresponding local region of the input face image and the registered face image, and the reciprocal of the sum may be used as the similarity. In this case, the face data 302 of the registration table 300 includes feature amounts (position information of face feature points, local regions in the vicinity of the face feature points) calculated for the registered face images in advance instead of the registered face images. Or the average value of the threshold values).

  When the face collating means 13 determines that the person to be collated is a registrant, the face collating means 13 sends a result notification indicating that the collation is successful to the voice output determining means 14, and when it is determined that the person to be collated is not a registrant, the result notification indicating that the collation is not successful. Is sent to the sound output determination means 14.

When the voice output determination unit 14 receives the result notification indicating the collation success or the collation failure from the face collation unit 13, the voice output determination unit 14 outputs a creation request for the output voice information indicating the voice phrase to be output to the voice output unit 3 as text data. The information is sent to the voice information creation means 15.
First, when the collation target person succeeds in collation, the face image authentication apparatus is installed when the phrase of the output voice output from the face image authentication apparatus 1 is the fixed phrase “Thank you for your work” and the personal name of the collation target person. An example in which a plurality of users pass through the passage will be described. FIG. 4A shows an example of the relationship between the input image and the output sound when a plurality of users pass through the passage where the face image authentication device 1 is installed in a discrete manner. In FIG. 4A, an image 400 is an input image at time t, an image 401 is an input image at time t + 1, an image 402 is an input image at time t + 2, and an image 403 is at time t +. 3 is an input image at time 3, and an image 404 is an input image at time t + 4. In the example shown in FIG. 4A, a person 410 (A officer) is shown in the input image 400 at time t. For this reason, the face image authentication apparatus performs a collation process at time t, and outputs an output voice “Thank you, A officer” according to the collation result from time t. At time t + 2, no voice is output because there is no person to be collated in the input image 402. Then, a new person 411 (Mr. B) is shown in the input image 403 at time t + 3. Therefore, the face image authentication apparatus performs a collation process at time t + 3, and outputs an output voice “Thank you, Mr. B” according to the collation result from time t + 3.

  On the other hand, FIG. 4B shows an example of the relationship between the input image and the output sound when a plurality of users continuously pass through the passage where the face image authentication device 1 is installed. In FIG. 4B, an image 420 is an input image at time t, an image 421 is an input image at time t + 1, an image 422 is an input image at time t + 2, and an image 423 is at time t +. 3 is an input image at time 3, and an image 424 is an input image at time t + 4. In the example shown in FIG. 4B, the person 410 (A officer) is shown in the input image 420 at time t. For this reason, the face image authentication apparatus performs a collation process at time t, and outputs an output voice “Thank you, A officer” according to the collation result from time t. On the other hand, the person 411 (Mr. B) is newly shown in the input image 421 at time t + 1. In this case, the face image authentication apparatus performs the collation process at time t + 1. However, since the voice is being output at time t + 1, the output voice “Thank you, Mr. B” cannot be output. Therefore, the face image authentication apparatus outputs the output voice “Thank you, Mr. B” from time t + 2 after the output of the output voice “Thank you, Mr. A” is completed. However, the person 411 (Mr. B) has already entered the room at time t + 3, and the voice output cannot be completed before the person 411 (Mr. B) enters the room. In addition, a person 412 (Mr. C) is newly shown in the input image 422 at time t + 2. In this case, the face image authentication apparatus performs the collation process at time t + 2, but since the voice is being output at time t + 2, the output voice “Thank you, Mr. C” cannot be output. Therefore, after the output of the output voice “Thank you, Mr. B” is completed, the face image authentication device outputs the output voice “Thank you, Mr. C” from time t + 4. However, the person 412 (Mr. C) has already entered the room at time t + 4, and the voice cannot be output before the person 412 (Mr. C) enters the room.

As described above, the voice output process generally requires a longer time than the collation process, and when a large number of users are collated in succession, the voice output to all of the users may not be in time. Therefore, when a plurality of users pass continuously, it is preferable that the output voice for the second and subsequent verification target persons is shorter than the output voice for the first verification target person.
Therefore, the audio output determination unit 14 of this embodiment determines whether or not the audio output unit 3 is outputting audio. The audio output determination unit 14 causes the output audio information creation unit 15 to create normal output audio information (hereinafter referred to as standard audio information) when the audio output unit 3 is not outputting audio. The standard voice information creation request is sent to the output voice information creation means 15. On the other hand, when the audio output unit 3 is outputting audio, the audio output determination unit 14 outputs output audio information (hereinafter referred to as shortened audio information) shorter than the standard audio information. In order to make it generate for 15, a request for creation of shortened voice information is sent to the output voice information creation means 15.
For example, in the example shown in FIG. 4B, the voice output unit 3 sequentially outputs the output voices “Thank you, A officer”, “Thank you, Mr. B”, “Thank you, Mr. C”. . The part of the output speech “Thank you for your work” is a fixed phrase. As shown in FIG. 4B, when the person to be collated continues, it is sufficient to output only the first person. Because it takes time to output, it is a cause that cannot output all of them. Therefore, the output voice information creating unit 15 of the present embodiment creates a text in which the fixed phrase is omitted from the standard voice information as shortened voice information.
Further, when the result notification received from the face collating means 13 indicates that the collation has failed, the person to be collated cannot be specified, and therefore the personal name cannot be included in the output voice information. Therefore, in this case, the audio output determination unit 14 causes the output audio information generation unit 15 to generate output audio information (hereinafter, referred to as audio information for unregistered users) consisting only of fixed phrases for unregistered persons. Therefore, a request for creating voice information for unregistered persons is sent to the output voice information creating means 15.

Further, the sound output determination means 14 determines whether or not the sound output unit 3 is outputting sound based on whether or not the output sound information is stored in the output table stored in the storage unit 5. That is, the sound output determination unit 14 determines that the sound output unit 3 is outputting sound when the output sound information is stored in the output table, and when the output sound information is not stored. It is determined that the audio output unit 3 is not outputting audio.
FIG. 5 is a schematic diagram of an output table stored in the storage unit 5. In the output table 500 shown in FIG. 5, one row corresponds to one person to be verified that is being tracked by the face tracking unit 12. Information in each row of the output table is added by the control means 16 when the collation process of the person to be collated is completed and the output voice information creation means 15 according to the collation result is created by the output voice information creation means 15. . The information in each row of the output table is deleted by the control means 16 when the audio signal synthesized from the output audio information has been output to the interface unit 4. Therefore, when the output audio information is not stored in the output table 500, it can be determined that the audio output unit 3 is not outputting audio.
As described above, the output voice information creating unit 15 creates a text in which the fixed phrase is omitted from the standard voice information as the shortened voice information. Therefore, in order to output the words shown in the abbreviated voice information as a voice, the words shown in the standard voice information are output immediately before that, or the words and phrases shown in the standard voice information are shortened. It is preferable that the words shown in the voice information are output as voice. Therefore, the audio output determination unit 14 of the present embodiment uses the standard when the output table 500 stores only audio information for unregistered persons, that is, when standard audio information or shortened audio information is not stored. A request for creating voice information is sent to the output voice information creating means 15.

The output voice information creating unit 15 creates output voice information in accordance with the creation request received from the voice output determining unit 14.
The information stored in the registration table and the fixed phrase table will be described below in order to describe the output sound information created by the output sound information creating means 15.
In the registration table 300 shown in FIG. 3, each column in the third column from the left shows the personal name 303 of the registrant. The personal name 303 is a registrant's real name, surname, common name, etc., and is stored as text data. When a surname is used as the personal name 303, the name may be recorded only when there are a plurality of registrants with the same surname.
In each column of the fourth column from the left, an attribute 304 of the registrant is shown. An attribute 304 is for identifying an employee, a trader, a customer, and the like. When the attribute 304 is an employee, the attribute 304 is further identified by a position such as an officer, a department manager, or a person in charge.
In each column of the fifth column from the left, a personal name time length 305 indicating an output time required for outputting the registrant's personal name 303 by voice is shown. For example, the personal name time length 305 can be obtained by creating voice data in which the personal name 303 is voiced and measuring the time required for voice playback. Alternatively, the personal name time length 305 may be determined by the number of characters when the personal name 303 is written in hiragana. Further, in order to facilitate processing, the output time required for outputting the longest personal name 303 by voice may be commonly applied as the personal name time length 305 of all registrants.

  In the registration table 300, for example, in the top row, the identification information 301 about the registrant is “1”, the personal name 303 is “A”, and the attribute 304 is “officer”. And the personal name time length 305 is 'tm1'. In the second line from the top, the identification information 301 for the registrant is “2”, the personal name 303 is “B”, the attribute 304 is “in charge”, and the individual It is shown that the nominal time length 305 is “tm2”. In the third line from the top, the identification information 301 for the registrant is “3”, the personal name 303 is “C”, the attribute 304 is “director”, and the individual It is indicated that the nominal time length 305 is “tm3”. In the fourth row from the top, the identification information 301 for the registrant is “4”, the personal name 303 is “D”, the attribute 304 is “customer”, and the personal It is indicated that the nominal time length 305 is “tm4”. In the fifth row from the top, the identification information 301 for the registrant is “5”, the personal name 303 is “E”, the attribute 304 is “trader”, and the personal It is indicated that the nominal time length 305 is “tm5”.

FIG. 6 is a schematic diagram of a fixed phrase table stored in the storage unit 5 that stores information on the fixed phrases of the voice output by the voice output unit 3. As shown in FIG. 6, a boilerplate type 601 is shown in each column at the left end of the boilerplate table 600. In the present embodiment, as the fixed phrase type 601, the authentication fixed phrase 611 that is a portion of the fixed phrase among the words of the output voice when the verification of the verification target person is successful, and the output voice when the verification of the verification target person fails The non-authenticated fixed phrase 612 that is a part of the fixed phrase and the honorific phrase 613 that is the part that follows the personal name of the person to be collated among the words of the output speech are registered. In this embodiment, the authentication phrase 611, the non-authentication phrase 612, and the honorific phrase 613 are managed as text data.
Each column in the second column from the left shows a phrase 602 registered as a fixed phrase. The phrase 602 is stored as text data. “Good morning”, “Thank you for your hard work”, “Thank you for your hard work”, and “Welcome” are registered as the phrases of the authentication phrase 611. For example, when the attribute 304 of the person to be collated is “customer”, “welcome” is used, and the attribute 304 of the person to be collated is “Executive Officer”, “Department Manager”, “Responsor”, or “Seller”. In any case, it is preferable to use “Good morning” in the morning, “Thank you very much” during the business hours in the afternoon, and “Good work after hours”.

In addition, “Thank you for visiting us”, “Please call us by extension phone” and “Please operate the card” are registered as phrases of the non-authenticated fixed phrase 612. For example, when the person to be verified is a customer, it is preferable to use “Thank you for visiting” or “If you are interested, please call us by extension”. On the other hand, when the face image authentication device 1 is installed in a place where only the employee is a collation target and the customer rarely comes and an authentication device using an ID card is provided separately from the face image authentication device 1 Is preferably "Please operate the card".
In addition, “san”, “officer”, “manager”, and “sama” are registered as honorific phrases 613. For example, “sama” is used when the person to be collated is a customer or a contractor, and “officer”, “manager”, and “san” are used according to the job title when the person to be collated is a customer. Is preferred. Alternatively, “san” may be used in common for all collation subjects in order to facilitate processing.

In each column of the third column from the left, a time length 603 indicating an output time required for outputting the phrase of each fixed phrase by voice is shown. For example, each time length 603 can be obtained by creating voice data in which the phrase of each fixed phrase is voiced and measuring the time required for voice playback. Alternatively, each time length 603 may be determined by the number of characters when the phrase is expressed in hiragana. In order to facilitate processing, the output time required to output the longest word / phrase in various types by voice may be commonly applied as the time length 603 of all words / phrases of that type. Hereinafter, the time length 603 of the authentication boilerplate 611 is referred to as an authentication boilerplate time length, the time length 603 of the non-authentication boilerplate 612 is referred to as a non-authentication boilerplate time length, and the time length 603 of the honorific phrase 613 is referred to as a honorific phrase time length. Called.
For example, for authentication boilerplate 611, the authentication boilerplate length of “Good morning” is 'tg1', the authentication boilerplate length of “Thank you” is “tg2”, and “Thank you very much” The authentication boilerplate time length is “tg3”, and the authentication boilerplate time length of “I welcome you” is “tg4”. For non-authenticated boilerplate 612, “Thank you for visiting”, the non-authenticated boilerplate time length is “tn1”, and “Please call us on the extension phone” unauthenticated boilerplate The time length is “tn2”, and the authentication boilerplate time length of “Please operate the card” is “tn3”. As for the honorific phrase 613, “san” 's honorific phrase length is “tr1”, “executive” honorific phrase duration is “tr2”, and “director” honorific phrase duration is “tr3”. It is shown that the honorific time length of “sama” is “tr4”.

  Note that the authentication boilerplate 611, the non-authentication boilerplate 612, and the honorific phrase 613 shown in FIG. 6 are only examples, and the words and phrases can be changed according to the installation location, operation policy, administrator preference, and the like of the face image authentication device 1. It may be changed or increased or decreased.

Hereinafter, the output voice information created by the output voice information creation unit 15 will be described. The output voice information creating unit 15 creates standard voice information, shortened voice information, or unregistered voice information as output voice information in accordance with the creation request received from the voice output determination unit 14. When creating the standard voice information, the output voice information creating unit 15 creates the output voice information from the authentication fixed phrase 611, the personal name 303, and the honorific phrase 613. For example, the standard voice information for the person to be verified who is identified as the registrant whose identification information 301 is “1” during the afternoon business hours is “Thank you, A officer”.
On the other hand, the output voice information creating means 15 creates the output voice information from only the personal name 303 and the honorific phrase 613 when creating the shortened voice information. For example, the information for abbreviated speech for the person to be verified who is identified as the registrant whose identification information 301 is “1” during the afternoon business hours is “A officer”. The shortened voice information is shorter than the standard voice information because the authentication phrase 611 is not present, and the voice output of the phrase indicated in the shortened voice information is shorter than the voice output of the word indicated in the standard voice information. Done on time.
In addition, when creating the voice information for unregistered persons, the output voice information creating unit 15 selects any one of the unauthenticated fixed phrases 612 and creates the output voice information. The phrase selected from the non-authentication fixed phrase 612 is determined according to, for example, the installation location, operation policy, and the like of the face image authentication apparatus 1.
When the output voice information creating unit 15 creates the output voice information, the output voice information creating unit 15 sends the created output voice information to the control unit 16.

  The control unit 16 stores the output voice information created by the output voice information creation unit 15 in the output table 500 of the storage unit 5. As described above, when the collation for a certain collation target person is completed, the voice output unit 3 may output a voice to the collation target person collated before that. In this case, the voice output unit 3 cannot output the voice for the next person to be collated unless the current voice output is completed. Therefore, when the control unit 16 acquires the output voice information from the output voice information creation unit 15, the control unit 16 temporarily stores the acquired output voice information in the output table 500 of the storage unit 5.

Hereinafter, the output table 500 illustrated in FIG. 5 will be described. As described above, in the output table 500, one line corresponds to one person to be verified that is being tracked by the face tracking unit 12. Hereinafter, the information shown in one line is referred to as notification information.
In each column of the leftmost column, an information number 501 is shown. The information number 501 is an identification number assigned to each notification information, that is, for each person to be verified. For example, the information number 501 is a continuous positive integer that is incremented each time notification information is added.
Each column in the second column from the left shows an information type 502. This information type 502 is information for distinguishing whether the output audio information is standard audio information, shortened audio information, or unregistered audio information.
Also, output audio information 503 is stored in each column of the third column from the left.
In addition, the output time 504 is shown in each column of the fourth column from the left. This output time 504 is the time required for outputting the words shown in the output audio information 503 as audio. That is, when the information type 502 is standard audio information, the output time 504 is the sum of the authentication fixed phrase time length, the personal name time length 305, and the honorific phrase time length. On the other hand, when the information type 502 is shortened voice information, the output time 504 is the sum of the individual name time length 305 and the honorific phrase time length. When the information type 502 is unregistered voice information, the output time 504 is a non-authentication fixed phrase time length.
Also, the creation time 505 is shown in each column of the fifth column from the left. The creation time 505 is the time when each piece of notification information is created and recorded in the output table 500.

  For example, when there is no notification information in the output table 500 and the information number 501 of the notification information existing immediately before is “260”, the identification information 301 is “1” during the afternoon business hours. Assume that the verification is successful for the registrant. In that case, since the notification information including the standard voice information or the shortened voice information does not exist in the output table 500, the voice output determination unit 14 determines that the identification information 301 is “1” in the output voice information creation unit 15. Send a request for creating standard audio information to the registrant. Upon receipt of the creation request from the voice output determination means 14, the output voice information creation means 15 reads the authentication phrase “Thank you very much”, the personal name “A” of the registrant whose identification information 301 is “1”, and the identification information. The standard voice information “Thank you, A officer” is created by combining the honorific phrase “executive” of the registrant with 301 of “1”. Upon receiving this standard voice information from the output voice information creation means 15, the control means 16 sets the information number 501 as “261”, the information type 502 as “standard voice information”, and the output voice in the output table 500. Notification information is added with the business information 503 as “Thank you, A officer”. Furthermore, the control means 16 outputs the notification format time length “tg2” of “Thank you” for the notification information output time 504, the personal name time length “tm1” of the personal name “A”, and the honorific phrase “officer”. Record the sum 'tg2 + tm1 + tr2' of honorific phrase time length 'tr2', and record the current time '11: 31: 48 'at creation time 505.

Further, from this state, when collation is successful for a registrant whose identification information 301 is “2”, the output table 500 includes notification information including standard audio information. A request for creating shortened voice information is sent to the information creating means 15 for a registrant whose identification information 301 is “2”. Upon receipt of the creation request from the voice output determination means 14, the output voice information creation means 15 sends the personal name “B” of the registrant whose identification information 301 is “2” and the registrant whose identification information 301 is “2”. The short voice information “Mr. B” is created in combination with the honorific phrase “Ms.”. Upon receiving this shortened voice information from the output voice information creating means 15, the control means 16 sets the information number 501 to “262”, the information type 502 as “shortened voice information” in the output table 500, and the output voice. Notification information in which the usage information 503 is “Mr. B” is added. Furthermore, the control means 16 outputs the sum “tm2 + tr1” of the personal name “B”, the personal name time length “tm2” of the personal name “B”, and the title time length “tr1” of the honorific phrase “san” in the output time 504 of the notification information. Record the current time “11:31:49” at the creation time 505.
Further, from this state, when the collation succeeds for the registrant whose identification information 301 is “3”, the control means 16 stores the information number in the output table 500 as in the case of the registrant whose identification information 301 is “2”. Notification information in which 501 is '263', the information type 502 is “shortened voice information”, and the output voice information 503 is “C section manager” is added. Further, the control means 16 outputs the sum “tm2 + tr1” of the personal name time length “tm3” of the personal name “C” and the title time length “tr3” of the honorific phrase “department” at the output time 504 of the notification information. Record the current time “11:31:50” at the creation time 505.

  Further, in this state, when collation of the person to be collated fails, the voice output determination unit 14 sends a voice information creation request for unregistered persons to the output voice information creation unit 15. When the creation request is received from the voice output determination means 14, the output voice information creation means 15 creates voice information “Thank you for visiting” for unregistered persons. Upon receiving this voice information for unregistered persons from the output voice information creating means 15, the control means 16 sets the information number 501 to “264” in the output table 500, and sets the information type 502 to “for voice for unregistered persons”. Information ”, and notification information for“ Thank you for visiting ”is added to the output audio information 503. Further, the control means 16 records the non-authenticated fixed phrase time length “tn1” of “Thank you for visiting” at the output time 504 of the notification information, and records the current time “11:31:51” at the creation time 505. To do.

  Further, the control means 16 controls the sound output by the sound output unit 3. When the sound output unit 3 is not outputting sound, the control means 16 reads the output sound information 503 stored in the output table 500 and sends it to the sound synthesizing means 17, which is generated from the output sound information 503. A voice signal is received from the voice synthesis means 17. Then, the control unit 16 outputs the received audio signal to the interface unit 4 and causes the audio output unit 3 to output the audio.

  At this time, the control means 16 reads the output audio information 503 stored in the output table 500 in order from the top. However, after the words shown in the standard voice information are output in voice, the words shown in the voice information for unregistered users are output in voice, and further the words shown in the shortened voice information are output in order. The output sound is inappropriate. For example, in the example shown in FIG. 4, if the verification of the person 411 (Mr. B) fails, the voices are output in the order of “Thank you, A officer”, “Thank you for visiting”, “Mr. C”. Non-authenticated boilerplate 612 is sandwiched between “Thank you so much” and “Mr. C”. Therefore, the control means 16 reads out the unregistered voice information after the output of the voice generated from all the standard voice information and the shortened voice information stored in the output table 500 is completed. . Further, since the information for unregistered speech is composed only of fixed phrases, when there are a plurality of unregistered speech information in the output table 500, it is redundant to output speech for each. For this reason, when there are a plurality of unregistered voice information in the output table 500, the control means 16 causes the voice synthesis means 17 to generate a voice signal for the unregistered voice information only once, and the interface unit 4 And the audio output unit 3 outputs the audio.

  When the output of the voice signal generated from the standard voice information or the shortened voice information is completed, the control unit 16 deletes the notification information corresponding to the standard voice information or the shortened voice information from the output table 500. In addition, when the output of the audio signal generated from the unregistered user sound information is completed, the control unit 16 deletes all the notification information including the unregistered user sound information from the output table 500.

  The voice synthesis unit 17 generates a voice signal from the output voice information 503 received from the control unit 16 by using a known voice synthesis technique, and sends it back to the control unit 16. For example, the speech synthesizer 17 performs speech synthesis using a corpus-based speech synthesis technique. In that case, an intonation database in which intonation data (for example, pitch frequency, power, phoneme duration, etc. for each phoneme) is stored for each word of a predetermined unit (for example, accent phrase) for a large number of sentences, and a speech element for each intonation and reading. A voice database in which pieces are stored is prepared in the storage unit 5 in advance. The speech synthesizer 17 first analyzes the output speech information 503 to identify a word and determine reading. Next, the speech synthesizing unit 17 searches the intonation database for a phrase that is the same as or most similar to the phrase in the output speech information 503 and determines the position of the intonation of the sentence. Further, the speech synthesizer 17 generates a speech signal by searching and connecting an appropriate speech segment from the speech database based on intonation and reading. Alternatively, the voice synthesizer 17 may generate a voice signal by generating an artificial voice from the output voice information 503 using a formant synthesis technique.

The operation of the notification information registration process by the face image authentication apparatus 1 to which the present invention is applied will be described below with reference to the flowchart shown in FIG. The operation flow described below is controlled by the processing unit 6. The operation described below is performed at every acquisition time of the input image (every frame).
First, the face detection unit 11 of the processing unit 6 acquires an input image in which a person to be verified is shown from the imaging unit 2 via the interface unit 4 (step S701). Then, the face detection unit 11 detects an input face area that is an area in which the face of the person to be collated is captured from the acquired input image, and creates an input face image (step S702). Then, the face detection unit 11 determines whether or not the input face area has been successfully detected (step S703). When the face detection unit 11 fails to detect the input face area, the processing unit 6 executes the processes after step S701 again.
On the other hand, when the face detection unit 11 succeeds in detecting the input face area, the face detection unit 11 notifies the face tracking unit 12 of information representing the input face area. The face tracking unit 12 performs a tracking process on the input face area extracted by the face detection unit 11, and associates the input face area in which the face of the same person is shown (step S704).

The following steps S705 to S711 are performed for each input face image that has been extracted by the face detection unit 11 and created from the input face region that is the target of the tracking process by the face tracking unit 12 and has not yet been collated. . The face collating unit 13 creates each registered face of the registration table 300 stored in the storage unit 5 for an input face image that has been created from the input face region that is the target of the tracking process by the face tracking unit 12 and has not yet been collated. The image is collated (step S705), and it is determined whether or not the image belongs to the same person (step S706).
If the face collating unit 13 determines that the person to be collated is not a registrant (No branch in step S706), the face collating unit 13 sends a result notification indicating a collation failure to the voice output determining unit 14. When the voice output determination unit 14 receives a result notification indicating a verification failure from the face verification unit 13, the voice output determination unit 14 sends a request for generating voice information for unregistered persons to the output voice information generation unit 15. When receiving the creation request for voice information for unregistered persons from the voice output determining means 14, the output voice information creating means 15 creates voice information for unregistered persons (step S707).
On the other hand, when the face collating unit 13 determines that the person to be collated is a registrant (Yes in step S706), the face collating unit 13 sends a result notification indicating the collation success to the voice output determining unit 14. When the voice output determination unit 14 receives a result notification indicating that the collation is successful from the face matching unit 13, the voice output determination unit 14 determines whether notification information including standard voice information or shortened voice information is stored in the output table 500. (Step S708). If neither the standard voice information nor the shortened voice information is stored in the output table 500 (Yes in step S708), the voice output determining unit 14 outputs a request for creating the standard voice information as the output voice. To the use information creation means 15. When receiving the request for creating standard audio information from the audio output determining unit 14, the output audio information creating unit 15 creates standard audio information (step S709). On the other hand, when either the standard voice information or the shortened voice information is stored in the output table 500 (No branch in step S708), the voice output determining unit 14 issues a request for creating the shortened voice information. This is sent to the output voice information creating means 15. When the output voice information creating unit 15 receives a request for creating shortened voice information from the voice output determining unit 14, the output voice information creating unit 15 creates shortened voice information (step S710).

When the output voice information creating unit 15 creates the output voice information 503, the output voice information creating unit 15 sends the created output voice information 503 to the control unit 16. Upon receiving the output voice information 503 from the output voice information creation means 15, the control means 16 generates an information number 501, calculates an output time 504, acquires a creation time 505, and outputs the output voice information 503. Notification information is created based on the information number 501, information type 502, output voice information 503, output time 504, and creation time 505, and added to the output table 500 (step S711).
When the control means 16 creates notification information and adds it to the output table 500, notification information registration processing is performed for all input face images that have been created from the input face area that is the target of tracking processing and have not yet been collated. It is determined whether or not. If there is an input face image for which notification information registration processing has not yet been performed, control returns to step S705, and the processing of steps S705 to S711 is repeated. On the other hand, when the notification information registration processing is performed for all the input face images, the processing unit 6 executes the processing from step S701 onward again.

The operation of the voice output process by the face image authentication apparatus 10 to which the present invention is applied will be described below with reference to the flowchart shown in FIG. The operation flow described below is controlled by the processing unit 6.
First, the control means 16 of the processing unit 6 determines whether notification information is stored in the output table 500 (step S801). When the notification information is not stored in the output table 500 (No branch in step S801), the speech synthesis unit 17 waits in step S801 until new notification information is stored in the output table 500. On the other hand, when the notification information is stored in the output table 500 (Yes in step S801), the control unit 16 includes standard audio information or shortened audio information as output audio information 503 in any of the notification information. It is determined whether it is included (step S802).

  If any of the notification information includes standard audio information or shortened audio information as the output audio information 503 (Yes in step S802), the control means 16 is stored at the head of the notification information. The output voice information 503 (standard voice information or shortened voice information) of the notification information is read (step S803). Then, the control unit 16 sends the read output voice information 503 to the voice synthesis unit 17. When the voice synthesis unit 17 receives the output voice information 503 from the control unit 16, the voice synthesis unit 17 generates a voice signal from the received output voice information 503 and sends it back to the control unit 16. When receiving the audio signal from the voice synthesizing unit 17, the control unit 16 outputs the received audio signal to the interface unit 4 and causes the audio output unit 3 to output the audio (step S804). When the output of the audio signal generated from the output audio information 503 is completed, the control unit 16 deletes the notification information corresponding to the output audio information 503 from the output table 500 (step S805). When the notification information is deleted from the output table 500, the control unit 16 executes the processing after step S801 again.

  On the other hand, when none of the notification information includes the standard voice information or the shortened voice information as the output voice information 503 (No in step S802), the control unit 16 is not registered as the output voice information 503. Out of the notification information including the audio information for users, the output audio information 503 (information for unregistered users) stored at the head is read (step S806). Then, the control unit 16 sends the read output voice information 503 to the voice synthesis unit 17. When the voice synthesis unit 17 receives the output voice information 503 from the control unit 16, the voice synthesis unit 17 generates a voice signal from the received output voice information 503 and sends it back to the control unit 16. Upon receiving the audio signal from the audio synthesizing unit 17, the control unit 16 outputs the received audio signal to the interface unit 4 and causes the audio output unit 3 to output the audio (step S807). When the output of the audio signal generated from the output audio information 503 is completed, the control unit 16 deletes all notification information including the unregistered user audio information from the output table 500 (step S808). When all the notification information including the information for voice for unregistered persons is deleted from the output table 500, the control unit 16 executes the processing after step S801 again.

Hereinafter, an operation of the face image authentication device 1 of the present embodiment when a plurality of users pass through a passage where the face image authentication device is installed will be described.
FIG. 9 shows an example of the relationship between the input image and the output sound when a plurality of users pass continuously through the passage where the face image authentication device 1 is installed. In FIG. 9, an image 900 is an input image at time t, an image 901 is an input image at time t + 1, an image 902 is an input image at time t + 2, and an image 903 is an input image at time t + 3. An image 904 is an input image at time t + 4. In the example shown in FIG. 9, a person 910 (A officer) is shown in the input image 900 at time t. Therefore, the face image authentication device 1 performs a collation process at time t, and outputs an output voice “Thank you, A officer” as a collation result from time t to time t + 1. On the other hand, a person 911 (Mr. B) is newly shown in the input image 911 at time t + 1. In this case, the face image authentication device 1 performs the collation process at time t + 1, and the output voice “Mr. B” is omitted at time t + 2 when the output of the output voice “Thank you, A officer” is completed. Is output. Furthermore, a new person 912 (Mr. C) is shown in the input image 902. In this case, the face image authentication device 1 performs the collation process at time t + 2, and outputs the output voice “Mr. C” from which “Thank you for your hard work” is omitted at time t + 3 when the output of the output voice “Mr. B” is completed. To do. As a result, the face image authentication apparatus 1 can complete the audio output until the person 911 (Mr. B) and the person 912 (Mr. C) enter the room.

  As described above, the face image authentication device 1 to which the present invention is applied is a collation target person and a registrant registered in advance in the input face image extracted from the input image obtained by photographing the face of the collation target person. Are the same person. Then, the face image authentication device 1 creates output audio information for the person to be collated, generates an audio signal from the output audio information, and outputs it. The face image authentication device 1 does not output voice information that is generated when the face image authentication device 1 is outputting sound when the verification of the person to be verified is completed. Make it shorter than the output audio information created in this case. Thereby, the face image authentication apparatus 1 can shorten the audio | voice output with respect to the collation target person collated after 2nd when there exist a some collation target person from a normal audio | voice. Therefore, the face image authenticating apparatus 1 can sequentially collate a plurality of collation target persons and optimize the output timing of the voices for each collation target person when outputting the voices for the respective collation target persons.

  The preferred embodiments of the present invention have been described above, but the present invention is not limited to these embodiments. For example, the data stored in the registration table, the fixed phrase table, and the output table may not be text data, but may be audio signals in a wav format, mp3 format, or the like created in advance. In that case, a voice signal is stored in the storage unit as a personal name of the registration table and a phrase of each fixed phrase of the fixed phrase table, and the output voice information creating means replaces the text data as the output voice information. Then, an audio signal is created by connecting these audio signals. On the other hand, the control means outputs the voice signal created by the output voice information creation means to the interface unit as it is without sending it to the voice synthesis means.

The shortened voice information may be any information as long as it is shorter than the standard voice information. For example, the short voice information may be a text in which the personal name and the honorific phrase are omitted from the standard voice information, that is, a text including only a fixed phrase. Alternatively, the phrase of the shortened voice information is the same as that of the standard voice information, the voice signal by the shortened voice information is temporally compressed using a known voice processing technique, and the output voice is uttered by the shortened voice information. The voice may be output in a short time by making the speed faster than the utterance speed of the output voice based on the standard voice information. For example, the speech synthesizer calculates a correlation between neighboring frame data using a synchronized overlap-add (SOLA) technique, and shifts to a portion having the largest correlation and performs a crossfade. Is compressed in time. Alternatively, the voice synthesizing unit may temporally compress the voice signal based on the shortened voice information using a PICOLA (Pointer Interval Controlled OverLap and Add) technique, a Phase Vocoder technique, or the like.
In this modified example, when the audio signal is stored in the storage unit as the personal name of the registration table and the phrase of each fixed phrase of the fixed phrase table, the control means is created by the output voice information creating means (connection) The voice signal based on the shortened voice information is sent to the voice synthesis means. Then, the voice synthesis unit compresses the received voice signal in time and sends it back to the control unit.

  In addition, the face matching means does not determine that the matching has failed in a single face matching process, but does not determine that the matching has succeeded (over a plurality of frames) in multiple face matching processes. In the case of a check or when a timeout occurs, it may be determined that the verification has failed. In that case, the processing of steps S707 to S711 shown in FIG. 7 is not performed for an input face image that has not been determined to be either collation success or collation failure. For this reason, the face collating unit does not send a result notification to the voice determining unit for an input face image that has not been determined to be either collation success or collation failure in step S706.

  Further, for example, the face image authentication device may be configured not to output a voice to a person to be collated who has been determined to have failed. In this case, the voice output determination means does not send a request for creating voice information for unregistered persons to the output voice information creation means even if a result notification indicating a matching failure is received from the face matching means. Therefore, in this case, there are two types of information of the output audio information: standard audio information and shortened audio information.

  In the present embodiment, the voice output determination unit determines whether or not the face image authentication apparatus outputs a voice when the result notification is received from the face image matching unit, that is, when the face matching unit completes the face matching process. Thus, an example of determining the type of output audio information to be generated by the output audio information generating means has been shown, but the present invention is not limited to this. The voice output determination means outputs the output voice information creation means to create depending on whether or not the face image authentication device outputs voice at another timing, for example, when the face matching means starts the face matching process. The type of audio information may be determined. In that case, the face matching unit sends a start notification to the voice output determining unit before starting the face matching process, and the voice output determining unit is outputting a voice when receiving the start notification. It is determined whether or not.

  In addition, the control means does not store the notification information in the order in which the notification information is created in the output table (the order in which the collation processing is performed), but for example, the notification information about the standard voice information or the shortened voice information. The order may be changed so as to be stored before the notification information about the voice information for unregistered persons. In that case, if the control means outputs the sound in the order of the notification information stored in the output table, the control means outputs the output sound based on the standard sound information or the shortened sound information before the output sound based on the unregistered sound information. be able to.

  In the present embodiment, the sound output determination means determines whether or not the face image authentication device is outputting sound based on whether or not standard sound information or shortened sound information is stored in the output table. However, the present invention is not limited to this. For example, the audio output determination unit may determine whether or not the face image authentication apparatus is outputting audio based on the time when audio output is started and the time required for audio output. FIG. 10 shows a schematic diagram of the output table in that case. As shown in FIG. 10, the output table 1000 in the present modification example shows an output start time 1005 instead of the creation time 505 in the output table 500 shown in FIG. 5. This output start time 1005 is the time when the control means outputs the audio signal based on the output audio information 1003 to the interface unit, and is recorded by the control means.

When the output of the audio signal by the output audio information 1003 to the interface unit is started, the control unit of the present modification records the time at the output start time 1005. Further, the control means sets the output start time 1005 for the notification information including the standard voice information newly stored from the state where the notification information including the standard voice information or the shortened voice information is not stored in the output table 1000. Recorded in the storage unit as the standard audio output start time.
As for the shortened sound information recorded after recording the standard sound information, the time when the output of the sound signal to the interface unit is started depends on the output time of the output sound information recorded before that time. Therefore, the output start time 1005 is left blank. Alternatively, the actual output time of the audio signal to the interface unit may be recorded as a follow-up result.
On the other hand, when the voice output determination unit receives the result notification indicating the verification success or the verification failure from the face verification unit, the voice output determination unit acquires the latest standard voice output start time from the recording unit. Further, the voice output determination means includes the output time 1004 of the notification information including the standard voice information corresponding to the standard voice output start time, and the shortened voice information created after the creation of the notification information including the standard voice information. The output time 1004 of all notification information is acquired. Then, the audio output determination means calculates the total sum of all the acquired output times 1004 (hereinafter referred to as the total output time).
For example, in the output table 1000 shown in FIG. 10, “11:31:48”, which is the output start time in the notification information whose information number 1001 is “261”, corresponds to the standard audio output start time, and the information number 1001 is 'Tg2 + tm1 + tr2' which is the output time 1004 in the notification information of '261', 'tm2 + tr1' which is the output time 1004 in the notification information whose information number 1001 is '262', and the information number 1001 is '263' The sum of “tm3 + tr3”, which is the output time 1004 in “notification information”, corresponds to the total output time.

The voice output determination means determines that the face image authentication device is outputting voice if the time when the result notification is received from the face image matching means is before the time when the total output time is added to the standard voice output start time. Then, a request for creating shortened voice information is sent to the output voice information creating means. On the other hand, if the time when the result notification is received from the face image matching means is after the time when the total output time is added to the standard sound output start time, the sound output determination means determines that the face image authentication device is not outputting sound. The standard voice information creation request is sent to the output voice information creation means. Alternatively, the voice output determination unit determines whether the time when the start notification instead of the result notification is received from the face image matching unit is before the time when the total output time is added to the standard voice output start time. It may be determined whether or not the authentication device is outputting sound.
Further, the control means generates from the standard voice information or the shortened voice information so that it can be determined whether or not the face image authentication device is outputting the voice based on the time when the voice output is started and the time required for the voice output. Even when the output of the voice signal is completed, the notification information corresponding to the standard voice information or the shortened voice information is not individually deleted from the output table 1000. When the output of all the voices generated from all the standard voice information and the shortened voice information in the output table 1000 is completed, the control means displays all the notification information including the standard voice information or the shortened voice information. delete. Alternatively, when the notification information including standard audio information is newly added to the output table 1000, the control unit displays the notification information including the standard audio information or the shortened audio information for which audio output has been completed so far. delete. The control means may also delete the standard audio output start time when deleting the notification information including the standard audio information or the shortened audio information.
Also in this modification, the face image authentication device can appropriately determine whether or not sound is being output, and is outputting sound depending on whether or not output sound information is stored in the output table. The same effect as in the case of determining whether or not there is obtained. In the case of this modification, the face image authentication device determines whether or not sound is being output according to the output time of each output sound information, and changes the length of the output sound. Therefore, even when the length of the output voice information changes for each verification target person, such as when the personal name of the verification target person is included in the output voice information, for example, the voice at the appropriate timing for each verification target person Can be output.

  Further, in the present embodiment, the face image authentication device has shown an example in which the voice corresponding to the collation result is output in the order in which the collation processing is performed on each collation target person, but the present invention is not limited to this. For example, priorities for the order of outputting voices may be determined in advance according to the attributes of the registrant, and the face image authentication apparatus may output the voices for each person to be collated in descending order of priority. In that case, for example, the priority of the registrant whose attribute is “customer” is set to “high”, the priority of the registrant whose attribute is “trader” is set to “medium”, and the attribute is “officer” , The priority of the registrant who is one of “director” and “in charge” (that is, “employee”) is set to “low”. This priority is stored in the storage unit in association with the attribute of the registrant. When storing the notification information in the output table, the control means notifies the standard voice information or the shortened voice information already stored in the output table from the creation time (or output start time) and output time of the output table. From the information, notification information for which voice output has not yet started is extracted. And a control means inserts notification information newly in the position before the notification information in which priority is lower than the notification information newly stored among the extracted notification information. And a control means outputs the audio | voice by the information for output audio | voices in order of the notification information stored in the output table. As a result, the control means can output the voices in the descending order of priority for the standard voice information or the shortened voice information stored in the output table and whose voice output has not started yet.

  FIGS. 11A and 11B are schematic diagrams of output tables for explaining a change in the order of audio output in accordance with the priority of the registrant. First, as shown in FIG. 11A, in the output table 1100, the notification information of the standard audio information whose information number 1101 is' 131 'and the creation time 1105 is '09: 15: 29', and information It is assumed that the notification information of the information for shortened voice whose number 1101 is “132” and the creation time 1105 is “09:15:30” is stored, and these attributes are “employee”. At this time, when the notification information of the information for shortened voice whose attribute is “customer”, information number 1101 is “133”, and creation time 1105 is “09:15:31” is newly created, control is performed. As shown in FIG. 11B, the means inserts the newly created notification information before the notification information whose information number 1101 is '132', which is “employee” whose priority is lower than “customer”. . Then, the control means reads out the output voice information in the order of the notification information stored in the output table 1100, causes the voice synthesis means to generate a voice signal, and outputs it to the interface unit. Note that since the notification information positioned at the head of the output table 1100 is assumed to be outputting the output sound for the output sound information, the newly created notification information is related to the high priority. First, it is arranged after the second. That is, in this example, the newly created notification information is inserted after the notification information whose information number 1101 is '131'.

When the order of audio output is changed according to the priority, the order in the output table 1100 is the order in which the notification information is created (the order in which the collation process is performed), and the value of the information number 1101 is changed. Good. In that case, the control means can output the voices based on the output voice information in the order of the information numbers 1101, not in the order stored in the output table 1100, so that the voices can be output in the descending order of priority.
As a result, the face image authentication apparatus can preferentially output the voice according to the collation result with respect to the customer or the like, and can control the order of the voice output finely.

  As described above, those skilled in the art can make various modifications in accordance with the embodiment to be implemented within the scope of the present invention.

DESCRIPTION OF SYMBOLS 1 Face image authentication apparatus 2 Imaging part 3 Audio | voice output part 4 Interface part 5 Memory | storage part 6 Processing part 11 Face detection means 12 Face tracking means 13 Face collation means 14 Audio | voice output determination means 15 Output voice information preparation means 16 Control means 17 Voice Synthetic means

Claims (4)

An imaging unit that sequentially acquires input images obtained by capturing the person to be verified;
A storage unit for storing a registered face image of the registrant in advance;
An audio output unit for outputting audio to the verification target person;
Face detection means for extracting an image of the face area of the person to be collated as an input face image each time the input image is acquired;
A face collating unit that collates the registered face image with the input face image and determines whether or not they are the same person;
Audio output determination means for determining whether the audio output unit is outputting audio;
Output sound information creating means for creating output sound information based on the determination result of the sound output determining means;
Voice synthesis means for synthesizing an output voice signal from the output voice information;
A face image authentication device having a control means for outputting the output audio signal from the audio output unit,
The output sound information creating means includes the output sound information as:
If the voice output unit is not outputting voice at the time of determination by the face matching means, create standard voice information,
The face image authentication apparatus characterized in that when the sound output unit is outputting sound at the time of the determination, shortened sound information shorter than the standard sound information is created. The storage unit further stores an authentication boilerplate and a personal name of the registrant in association with the registered face image of the registrant,
The output audio information creating means includes:
The standard voice information is created from the authentication boilerplate and the personal name of the registrant determined to be the same person as the person to be collated in the input face image by the face collating means,
2. The face image according to claim 1, wherein the shortened voice information is created from the personal name of the registrant who is determined to be the same person as the person to be collated in the input face image by the face collating unit. Authentication device. The control means includes
Each time the output sound information creating means creates the output sound information, the output sound information is stored in the storage unit, while the sound output unit completes sound output of the output sound signal. Deleting the output audio information corresponding to the output audio signal from the storage unit;
The sound output determination means is
If the output audio information is stored in the storage unit, the audio output unit determines that the audio output unit is outputting audio, and if the output audio information is not stored in the storage unit, the audio output unit The face image authentication device according to claim 1, wherein it is determined that is not outputting voice. In the storage unit, an attribute is stored for each registrant, and a priority that predetermines the order in which the control unit outputs the output audio signal to the audio output unit is determined in accordance with the attribute. Degree is remembered,
The control means outputs the output audio signals for the output audio information for which the audio output by the audio output unit has not started among the output audio information stored in the storage unit in descending order of priority. The face image authentication device according to claim 3, wherein the voice output unit outputs a voice.

Images