JPH06119433A - Figure authentication device - Google Patents

Abstract

PURPOSE:To provide the figure authentication device which does not strict the posture of a user at the time of authenticating a figure by separating a background area and a face area from plural image data and measuring a decided feature amount from the face area. CONSTITUTION:The reference feature amount of the posture or the expression is calculated with a decided evaluation formula by an area processing part 5 and held in a feature amount recording part 9 as a value for judgement. At the time of authentication, the measured feature amount is calculated by an authentication part 7 while using the same evaluation formula and compared with the recorded value for judgement. At such a time, when the relation between two data satisfies decided conditions, it can be judged that the image data are effective as image data for authentication from the viewpoint of the posture or the expression. Namely, the authentication of the figure is not performed with one piece of image data at certain time but performed by inputting plural image data for certain fixed time and performed by using the image data satisfying the previously decided conditions.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention inputs a face image of a person,
The present invention relates to a person authentication device that measures the characteristic amount and performs authentication.

[0002]

2. Description of the Related Art Authentication of a person is considered as a security measure, and a system using a face image is regarded as promising because it does not burden a user.

Conventionally, the face area and the background area are separated from the input image data, the characteristic amount of the predetermined face image data is measured, and the characteristic amount is compared with the reference characteristic amount registered in the apparatus. Authenticated the person. However, at present, there are the following problems.

1) Due to the difference between the condition of photographing the image data for which the reference feature amount is calculated and the condition of photographing the image data for authentication (background, lighting condition, distance, camera angle, etc.), the authentication accuracy rate is It depends a lot. 2) Even if the conditions at the time of capturing image data are the same, the feature amount obtained differs depending on the posture of the person, and therefore the authentication accuracy rate is greatly affected.

3) Even if the position or state of the face is limited to the front face or the side face, the feature amount obtained by the change in the facial expression (closed eyes, laughing, wide open mouth) Since they are different, the accuracy rate of authentication is greatly affected. Conventionally, the following solutions have been considered for these problems. The problem of No. 1 can be solved to some extent by photographing the reference image and calculating the feature amount under the condition that the authentication is actually performed.

The second problem is that authentication is performed by limiting the posture of the person to the front face or the side face so that such an error does not occur. Recently, postures are classified, face images corresponding to each posture are prepared, the feature amount of each image is calculated and registered in the device as a reference value, and the registered values of all postures are compared at the time of authentication to perform authentication. A device for performing is proposed. Further, there is also devised a device that determines the posture of a face from the positions of eyes and mouth, performs linear conversion on the face image data to form a front face, calculates a feature amount from the obtained image, and performs authentication.

For problem 3, the face expressions are classified in the same manner as problem 2, and a reference image corresponding to each expression is prepared.
A device is conceivable in which the characteristic amount of each image is calculated and registered in the device as a reference value, and at the time of authentication, the registered values of all expressions are compared to perform authentication.

However, it has been said that it is a good device to restrain the posture and expression of the user at the time of authentication from the viewpoint of the human interface because it imposes a burden on the user.

Further, when a plurality of reference images are created, the characteristic amount of each data is calculated and recorded in the apparatus, there arises a problem that the amount of data to be recorded and the amount of calculation at the time of authentication increase. . Further, the method of detecting the state of the face and performing the linear conversion may cause a calculation error during the conversion. Considering the above, it is considered that there is a problem in performing person authentication with one piece of image data.

[0010]

As described above, according to the conventional method, the posture and facial expression of the user must be restricted at the time of authentication in order to avoid the problem. A device was also devised in which a plurality of reference images were prepared, and the feature amount of each image was calculated and registered. However, the feature amount of images of all postures and expressions is registered and compared. There is a problem that the amount of data to be increased and the amount of calculation are increased. An object of the present invention is to provide a person authentication device that does not restrain the posture of the user when authenticating a person and does not increase the amount of data to be recorded or the amount of calculation.

[0011]

The present invention is characterized by performing authentication from a plurality of image data in order to achieve the object. Therefore, in the present invention, a means for separating the background area and the face area from the input image data, a means for measuring the characteristic amount determined from the face area, and the obtained reference characteristic amount are used in advance. Whether the input image data satisfies the specified conditions by calculating the specified evaluation formula, holding the reference data for judgment, and comparing the evaluation formula value with the reference data. It is characterized in that it is provided with means for making a judgment. Further, it is provided with means for performing authentication using the obtained feature amount if the input image is determined to be valid, and means for inputting image data again if it is determined to be invalid. And

[0012]

In the present invention, when the reference characteristic amount is calculated, the characteristic amount of the posture or facial expression is calculated by a predetermined evaluation formula, and is stored in the device as a judgment value. At the time of authentication, the measured feature amount is calculated using the same evaluation formula, and is compared with the determination value recorded in the device. At this time, if the relationship between the two data satisfies a predetermined condition, it can be determined that the image data is effective as image data for authentication in terms of posture and expression. That is, a device having a high authentication rate is provided by using a means for obtaining the same image data as when the reference feature amount was calculated. At this time, if it is determined that the image data is invalid, the image data is discarded and new image data is input. By repeating this process until effective image data is obtained, it is possible to prevent the change of the feature amount due to the posture and change of the face.

That is, the person is not authenticated with one piece of image data at a certain time, but a plurality of pieces of image data are input for a certain fixed time, and the image data satisfying a predetermined condition is input. Use to authenticate. Therefore, the user does not have to match the device at the time of authentication, but the device automatically selects the image data for authentication, so that the device with less burden on the user can be provided.

[0014]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an example of the configuration of the present invention.

The analog signal input to the camera is converted into a digital signal by the image input section 1 and recorded in the memory 2. Next, the face area detection unit 3 cuts out the area shown in FIG. 2 by performing a predetermined calculation on the entire image of the memory 2 and records it in the memory 4.

The area processing unit 5 calculates validity determination data by performing a predetermined calculation on the face area data and sends the validity determination data to the validity determination unit 6. In addition, the feature amount indicating the feature of the area is calculated and sent to the authentication unit 7.

The validity determining unit 6 performs a predetermined calculation using the data from the area processing unit 5, compares the result with a predetermined value, and the input image is effective for authentication. It is determined whether or not the result is sent to the control unit 8. The control unit 8 generates an authentication valid signal to the authentication unit 7 when it is valid, and generates an image input instruction signal to the image input unit 1 when it is invalid.

The authentication section 7 determines the person by calculating the characteristic amount of the area and the reference value recorded in the characteristic amount recording section 9. The control unit 8 uses the result as an index signal.
Send to.

The control unit 8 is composed of a CPU, a ROM, and a RAM, and generates a signal necessary for operating this device by following a predetermined processing procedure. Further, a necessary signal is sent to the authentication result display unit 10 based on the index signal sent from the authentication unit 7. The authentication result display unit 10
It has a means for displaying the result based on the signal sent. Next, the processing contents and processing means of each block will be described in detail below. FIG. 3 is a block diagram showing an example of the configuration of the image input unit 1.

The image input section is composed of an A / D conversion section 11, a matrix section 12, and a memory 2 (the same as in FIG. 1). When an input instruction signal from the control section is detected, an analog signal from the camera is detected. It is converted into a digital signal for a frame and recorded in the memory 2. At this time, the analog RGB input signal from the camera is converted into a digital signal by the A / D conversion unit 11, and the matrix unit uses the following equation to obtain:
The Y, RY and BY signals are converted and recorded in the memory 2. Y = 0.299 * R + 0.584 * G + 0.117 * B RY = R−Y BY = B−Y FIG. 4 is a block diagram showing an example of the configuration of the face area detection unit 3.

In this example, the face area detecting section is composed of a polar coordinate converting section 21, a pixel validity determining section 22, a face area determining section 23, and memories 2, 4 and 24. (Memory 2 and 4 are shown in Figure 1.
Same as

The polar coordinate conversion unit 21 performs the following processing from the RY signal and the BY signal recorded in the memory 2 to create a PHS signal, and the pixel validity determination unit 22 uses the PHS signal, RY signal, and BY signal to determine the effective pixel. Is determined. PHS = ATAN (BY / RY) RY! = 0 PHS is defined as a signal representing a hue, and if the RY signal is 0, no conversion is performed. The pixel validity determination unit 22 determines validity / invalidity using the following rules, and records the determination result in the memory 2. 1 If the value of the luminance signal is smaller than a predetermined value, it is regarded as an invalid pixel. 2 When the RY signal is 0, the pixel is invalid. 3 If the PHS signal is not a value within a predetermined range, it is regarded as an invalid pixel. The face area determination unit 23 examines the data in the memory 4 and extracts an area satisfying a predetermined size as a face area.

At this time, the coordinate value of the face area and the data of the width and the height are sent to the validity judging section 6. Using this data, the data corresponding to the face area is extracted from the memory 2 and recorded in the memory 24. 5 to 6 are block diagrams showing examples of the configuration of the area processing unit 5.

FIG. 5 shows an example of the configuration of the area processing unit of the apparatus for detecting a plurality of predetermined areas (N pieces) from the face area and performing authentication. The N area processing units 31 are
In each case, a predetermined area is detected in accordance with the predetermined processing, and the processing is performed for each area to detect the judgment data and the feature amount. At this time, each processing unit receives the data of the face area of the memory 24 as input, sends the determination data obtained by calculation to the validity determination unit 6, and sends the feature amount to the authentication unit 7.

FIG. 6 shows a more detailed block diagram of the area processing unit, taking the area processing for the left eye as an example. At this time, the region processing unit includes the left eye region detection unit 32, the normalization unit 33, and the linear conversion unit 3.
4. Comprised of a calculation unit 35.

The left-eye area detection unit detects the left-eye area from the face area of the memory 24 according to a predetermined process, and determines whether the coordinate value, width, and height of the area are valid data. Send to Part 6. The normalization unit 33
The image data of the detected left-eye area is normalized based on the normalized reference data sent from the validity determining unit 6 to the area size. Further, the data is adjusted so that the average value of the luminance signal of the area becomes a predetermined value and the maximum value and the minimum value of the data also become the predetermined values.

The linear conversion section 34 performs a linear conversion on the normalized data to convert it into another space. Examples of linear transformation include DFFT and DCT, which transform data into a two-dimensional frequency space. Arithmetic unit 3
5 selects a predetermined coefficient and sends it to the authentication unit 7 as a feature amount. FIG. 7 shows an example of selection.

FIG. 8 shows a method of using a predetermined feature point instead of the area. In this example, the mouth area detection unit 4
The position information of the mouth area obtained in 1 is used to perform a search in the horizontal and horizontal directions from the center coordinates of the mouth, and the cheek position detection unit 42
To detect the cheek edge. Then, the distance from the center point of the mouth is normalized using the reference data from the validity determining unit 43.
Normalize with. The data is sent to the validity judging section as judgment data and is also sent to the authentication section as a feature amount. FIG. 9 is a block diagram showing an example of the configuration of the validity determination unit 6. The validity determination unit is composed of a normalization processing unit 51, a validity determination processing unit 52, and a reference data recording unit 53.

The normalization processing section 51 performs a predetermined calculation on the judgment data sent from each area, and obtains the condition in order to match the condition of the input image data with the condition when the reference value is created. The reference data for normalization is created based on the obtained data, and the judgment data is normalized.

The validity determination processing unit 52 determines whether the input image data is valid or invalid by comparing the determination reference value of the reference data recording unit 53 with the normalized determination data. The determination result is sent to the control unit 9. less than,
An example of an algorithm for performing validity determination processing will be described.

The left eye area, the right eye area, and the mouth area are used as the areas, the left cheek edge and the right cheek edge are used as the characteristic points, and the validity / invalidity determination method is performed using the positional information of each area and the height width data of the area. An example is shown. 10 and 11 are diagrams showing the names of the reference data. First, the processes 1 to 4 are performed to create and normalize the reference data. 1 Find the center point of each area eye_center_x, eye_center
_Y reye_center_x, reye_center
_Y mount_center_x, mouth_cent
or_y 2 Measurement of the distance from the left edge of the mouth region to the left cheek Measurement of the distance from the right end of the mouth region to the right cheek The position is determined, and the value between both areas is used as the normalization reference data. face_center_x = (eye_center_x + reye_center_x) / 2 face_center_y = (eye_center_y + reye_center_y) / 2 Perform normalization data for normalization using 4 base_width_reye_center_y = (eye_center_y + rye_center_y) / 2 base_width = reye_center_y

Next, a predetermined calculation is performed on the normalized judgment data, and when the predetermined reference value is satisfied, the judgment is made valid, and otherwise, the judgment is made invalid. An example of the judgment calculation is shown below. 1 y coordinate of the center point of the left eye area and y of the center point of the right eye area
Calculate the difference in coordinates and invalidate if it exceeds a predetermined value. 2 Calculate the difference between the distance from the left edge of the mouth area to the left cheek and the distance from the right edge of the mouth area to the right cheek, and invalidate if it exceeds a predetermined value. 3 Calculate the difference between the x-coordinate of the center point of the face and the x-coordinate of the center point of the mouth, and invalidate if it exceeds a predetermined value. FIG. 12 is a block diagram showing an example of the configuration of the authentication unit 7. The authentication unit 7 includes a weighting processing unit 61, an authentication processing unit 62, and a reference feature amount recording unit 63. The authenticating unit 7 checks the valid signal from the control unit, and if the valid signal is valid, the authenticating unit 7 performs the authenticating operation in the following procedure.

The feature quantities sent from each area processing section and the feature point processing section are weighted in advance by the weighting processing section 61, and then combined into one authentication vector. The authentication processing unit 12 calculates the detected authentication vector and the reference vector recorded in the reference feature amount recording unit 63.

As a calculation means, a method using the squared error of each vector value can be considered. After the calculation with all the reference vectors is completed, the index signal indicating the reference vector showing the minimum value is sent to the control unit 8.

At this time, the calculation result is the reference feature amount recording unit 63.
When the validity determination reference value recorded in No. 3 is not satisfied, a predetermined signal is sent to the control unit 8 as an index signal. FIG. 13 shows the configuration of the authentication result display section.

Based on the index signal sent from the authentication section 7, the control section 8 determines the true or false of the person from the recorded correspondence data between the index and the person. Then, the necessary data is sent to the display section. At this time, if the index signal has a predetermined value, it is determined that the authentication has failed, and the data is sent to the display unit according to a predetermined instruction. Next, another embodiment for realizing the present invention will be described.

In the above embodiment, the coefficient in the frequency space of each area is used as the feature amount, but a system using the position coordinates, width and height of each area is also conceivable. FIG. 14 shows a block diagram of an area processing unit for performing the processing.

In the above embodiment, the face area is detected by using the color as a clue, but the difference data is calculated from a plurality of input image data and binarized to detect an area having a certain size. A method is also possible.

Further, the above-described embodiment is a system in which an effective frame is detected on the time axis and, if it is determined that the frame is an effective frame, the image at that time is used as an authentication image for authentication. However, as another embodiment, the valid / invalid determination is performed for each area, the feature amount at that time is held for the valid area, no further processing is performed, and the image input is performed until the invalid area disappears. If all areas are validated by performing the above, an apparatus that performs authentication by using the retained feature amount may be considered. As another embodiment, a device in which a frame in which the eye region is effective and a frame in which the mouth region is effective are vertically combined and used as one input image can be considered.

In the above embodiment, a plurality of non-linear processings are performed at each processing stage, but in some cases the predetermined conditions may not be satisfied. As shown in the block diagram of FIG. 15, the effective range is determined in advance for the numerical values and the characteristic amounts obtained after each processing, and if a numerical value exceeding the range is detected, it is determined that the processing has failed. An embodiment having a means for generating an invalid signal in the validity judging section is also conceivable.

[0041]

According to the person authenticating apparatus of the present invention, as compared with the conventional person authenticating apparatus, the user is not required to take a posture or a facial expression, so that the person authenticating apparatus is excellent. Further, it is a device in which the authentication rate does not decrease due to changes in the posture and facial expressions of the user. It is expected that security will be strengthened in the advanced information society in the future, and a more flexible system will be required. The present invention is effective in such a system.

[Brief description of drawings]

FIG. 1 is a block diagram showing the overall configuration of an embodiment of the present invention.

FIG. 2 is a diagram showing how a face area is cut out from image data.

FIG. 3 is a diagram showing a configuration of an image input unit.

FIG. 4 is a diagram showing an example of a configuration of a face area detection unit.

FIG. 5 is a block diagram showing a configuration of an area processing unit.

FIG. 6 is a more detailed block diagram of an area processing unit taking the left eye area processing as an example.

FIG. 7 is a diagram showing how a coefficient is selected as a feature amount.

FIG. 8 is a diagram showing a method of calculating a feature amount using position information of a mouth region.

FIG. 9 is a block diagram showing a configuration of a validity determination unit.

FIG. 10 is a diagram showing names of reference data of a face area.

FIG. 11 is another diagram showing names of reference data of a face area.

FIG. 12 is a block diagram showing a configuration of an authentication unit.

FIG. 13 is a block diagram showing the configuration of an authentication result display unit.

FIG. 14 is a block diagram showing another example of a region processing unit.

FIG. 15 is a block diagram showing the overall configuration of another embodiment of the present invention.

[Explanation of symbols]

 1 ... Image input section 2, 4 ... Memory 3 ... Face area detection section 5 ... Area processing section 6 ... Validity determination section 7 ... Authentication section 8 ... Control section 9 ... Feature amount recording section 10 ... Authentication result display section

Claims (3)

[Claims] 1. An input means for inputting image data, a storage means for holding a reference feature amount used for authentication of a person, and a separating means for separating a face area and a background area from the input image data. A measuring means for measuring a predetermined characteristic amount from the separated face area; an authenticating means for authenticating a person by comparing the measured characteristic amount with a reference characteristic amount; and an output means for outputting an authentication result. A person authentication device comprising: 2. A means for holding a value for judging the validity of the measured characteristic amount, a means for performing a predetermined arithmetic operation on the measured characteristic amount, and a value obtained by the arithmetic operation. A means for comparing the value for judgment and judging the validity of the image data is further provided.
Person authentication device described. 3. When the image data is determined to be invalid, the authentication means does not compare the measured feature quantity with the reference feature quantity, and the input means inputs the image data again. Item 2. The person authentication device according to item 2.