KR101189765B1 - Method and apparatus for classification sex-gender based on voice and video - Google Patents

Abstract

The present invention relates to a method and apparatus for identifying a gender and age of a specific person from input image information and audio information. Specifically, a combination of voice recognition and face recognition is considered in consideration of the correlation between gender information and age information. The present invention relates to an identification device and a method for accurately calculating gender and age.

Gender-age identification method according to the invention, the step of collecting the image information and audio information; Extracting one or more feature values with respect to the collected voice information, and determining gender and age using voice information for determining gender and age using the extracted feature values; Extracting one or more feature values with respect to the collected image information, and determining gender and age using face information for determining gender and age using the extracted feature values; And finally determining the gender and age by combining a gender and an age determined using the voice information and a gender and an age determined using the face information.

Face recognition, voice recognition, gender identification, age identification, intelligent DB

Description

Gender-age discrimination method and device based on voice and video {METHOD AND APPARATUS FOR CLASSIFICATION SEX-GENDER BASED ON VOICE AND VIDEO}

The present invention relates to a method and apparatus for identifying a gender and age of a specific person from input image information and audio information. Specifically, a combination of voice recognition and face recognition is considered in consideration of the correlation between gender information and age information. The present invention relates to an identification device and a method for accurately calculating gender and age.

The present invention is derived from the research conducted as part of the IT source technology development project of the Ministry of Knowledge Economy and the Ministry of Information and Telecommunication Research and Development. Technology development].

As the gender and age discrimination technology of the user according to the prior art, there is a method using a personal identification means such as an electronic resident's card, a method using a face recognition, a method using a voice recognition and the like.

As a method of using the personal identification means, the age recognition method using the electronic resident ID (Korean Patent Publication No. 1999-0008679) has the inconvenience that each individual must always carry a personal identification means such as electronic resident ID, Personal identification means such as resident card, there is a problem that can be easily lost, damaged, counterfeited.

In the face recognition method used as a conventional gender-age discrimination technique, since the gender and age are determined only by the face image information, it is difficult to reflect the characteristics for each individual, so that the recognition accuracy is low, and the recognition method using the voice recognition In the case of, since gender and age are determined only by voice information, there is a problem in that the recognition accuracy is lowered when the voice characteristics such as women and children are similar.

In addition, the conventional method of discriminating face or voice based recognition does not discriminate age and sex by reflecting specificity in which feature distribution varies according to gender or specificity in which sex feature distribution varies according to age. The disadvantage is that the amount of computation is large.

The present invention is to solve the above-mentioned problems, the problem to be solved by the present invention is to use the interrelationship of gender information and age information, and gender-age that can improve the recognition accuracy by combining voice recognition and face recognition An identification method and an apparatus thereof are provided.

Gender-age identification method according to an aspect of the present invention, collecting the image information and audio information; Extracting one or more feature values with respect to the collected voice information, and determining gender and age using voice information for determining gender and age using the extracted feature values; Extracting one or more feature values with respect to the collected image information, and determining gender and age using face information for determining gender and age using the extracted feature values; And finally determining the gender and age by combining a gender and an age determined using the voice information and a gender and an age determined using the face information.

According to another aspect of the present invention, a gender-age identification apparatus includes an input unit for collecting image information and audio information; A voice processor extracting a feature value with respect to the collected voice information and determining a gender and an age from the voice information using the extracted feature value; An image processor extracting a feature value with respect to the collected image information and determining a gender and an age from the image information using the extracted feature value; And a final determination unit which finally determines the gender and age of the specific person by combining a gender and age determined by the image processor and the gender and age determined by the voice processor.

Since the present invention is performed by combining the voice recognition and the face recognition, the recognition accuracy is improved compared to the conventional method using only the voice recognition or the method using only the face recognition.

In addition, the present invention, because the correlation between the gender information and age information, for example, age discrimination recognizes the age and gender by reflecting the specificity of the distribution of the characteristics according to gender or the specificity of the distribution of gender characteristics according to age, etc. Compared with the recognition method, high accuracy can be guaranteed.

In addition, the present invention primarily groups voice information on the basis of the characteristics that can be easily distinguished for each input information with respect to the input information in the feature extraction, and the characteristics of each group for each group distinguished according to the criteria. By using the method of extracting the feature value by reflecting, it is possible to secure the accuracy of identification and to perform the quick identification by eliminating the redundancy of the operation.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an embodiment of a gender-age identification apparatus according to the present invention.

As shown in FIG. 1, the gender-age identification apparatus according to the present invention may include an input unit 10, an age-gender calculator 20, and an output unit 30.

The input unit 10 collects video information and audio information of a specific person.

The input unit 10 may include image information acquisition means such as a camera capable of acquiring image information, and acoustic information acquisition means such as a speaker capable of acquiring sound information.

In addition, the input unit 10 may separately extract only the voice information of the specific person from the face extracting means for extracting only the face information of the specific person separately from the image information acquired by the image information acquiring means and the sound information acquired by the sound information acquiring means. It can be configured to include a voice extraction means. In this case, since each feature extracting means of the age-gender calculation unit 20 does not need to extract face information and voice information from image information and sound information each time, it is possible to perform fast calculation.

The implementation of such face extracting means and speech extracting means may be implemented using conventional face detection techniques. For example, for face extraction, knowledge-based methods, feature-based methods, template-matching methods, appearance-based methods, thermal infrared ( Infra Red), a three-dimensional face recognition method, a multi-modal method can be used to implement the face extraction means.

The age-gender calculator 20 includes a voice processor 100 for determining age and gender based on voice information, an image processor 200 for determining age and gender based on image information, a voice processor 100 and an image processor. The final determination unit 300 is configured to determine the age and gender by combining the calculation results of the 200.

The output unit 30 outputs the age and gender transmitted from the age-gender calculator 20.

Hereinafter, the age-gender calculator 20 will be described in detail with reference to FIGS. 2 and 3.

FIG. 2 is a detailed configuration diagram of the voice processing unit 100 shown in FIG. 1.

As shown in FIG. 2, the voice processing unit 100 includes a voice feature extraction unit 110 for extracting feature values from voice information and a voice operator 120 for determining gender and age from the extracted feature values. Can be configured.

In more detail, the voice feature extractor 110 extracts one or more feature values or feature vectors (hereinafter, referred to as "feature values") with respect to the voice information. The voice feature extractor 110 includes a linear predictive coefficient method, a cepstrum method, a mel frequency cepstral coefficient (MFCC) method, and an energy per frequency band (Filter Bank Energy). Method) or a combination thereof to extract feature values.

The voice feature extractor 110 extracts a plurality of feature values from the same voice information by applying a plurality of feature value determination methods described above, or uses a single feature value determination method to generate a plurality of feature values using a plurality of samples. When the feature values are obtained for M speech samples using N feature discrimination methods, the feature values may be represented in a matrix form of (N * M).

In this embodiment, in order to accurately and quickly perform the feature extraction for voice, gender feature extraction unit 111, age-specific feature extraction unit-M (112), age-specific feature extraction unit -FC 113, age-specific feature extraction unit- The voice feature extractor 110 is configured to include an F 114 and a gender feature extractor-C 115.

The sex feature extracting unit 111 extracts a feature value by reflecting the difference between the male and the female, that is, the sex feature, with respect to the input voice information, and extracts the voice information based on the extracted feature value from the male group M or the female. And children's groups (FC).

The age-specific feature extractor-M 112 extracts feature values with respect to the voice information divided into the male group M by the gender feature extractor 111, and in this case, the input voice information is determined as the voice information of the male. Since it is the voice information, the feature value can be extracted by reflecting the age-specific features of the male.

The age-specific feature extractor-FC 113 may extract feature values by reflecting age-specific features of women and children with respect to voice information divided into women and children groups (FC) by the sex feature extractor 111. As a result, the input voice information is divided into the female group (F) and the children group (C). At this time, the children's group (C) is a group for people before the metamorphosis difficult to distinguish the gender characteristics.

The age-specific feature extractor-F 114 may extract feature values by reflecting the age-specific features of women with respect to the voice information divided into the female group F by the age-specific feature extractor-FC 113.

The sex feature extractor-C 115 extracts a feature value by reflecting the sex feature of the child with respect to the voice information classified into the child group C by the feature extractor-FC 113 for each age.

The voice operator 120 may receive the feature value extracted by the voice feature extractor 110 and determine the gender and age of the input voice as described above.

To this end, the speech operation unit 120 based on the combinational operation unit that determines the representative feature value by reflecting the weight with respect to the feature value extracted by the speech feature extractor 110 and the determined representative feature value, the reference feature value for each gender and age Or a determination unit for determining gender and age by referring to a reference DB storing audio and image reference samples.

In addition, as shown in FIG. 2, the voice operation unit 120 is a combination operation unit optimized for the male group M, the female group F, and the child group C grouped by the voice feature extraction unit 110, respectively. And a discriminating unit, respectively.

Hereinafter, a description will be given with reference to the embodiment shown in FIG.

The voice operator 120 receives a feature value extracted from voice information divided into a male group M from the voice feature extractor 110 and calculates a gender and an age. Voice operation unit that calculates gender and age by receiving feature values extracted from voice information divided by) -F (122) and the feature value extracted from voice information divided by children group (C). It may be composed of a speech operation unit -C (123) to calculate.

In more detail, the voice operator-M 121 includes a combination operator-M 121A and a discriminator-M 121B, and the combination operator-M 121A includes voice information divided into a male group M. Determining the representative feature value by weighting one or more feature values extracted in the determination unit, the determination unit -M (121B) can determine the gender and age with reference to the reference DB based on the representative feature value. In addition, since the combination operation unit-M 121A receives voice information divided into male groups and performs combination operation, the feature extracted from the gender feature extractor 111 and the age-specific feature extractor-M 112 according to the above description. You can receive values.

Similarly, the speech computing unit -F 122, the combinational computing unit -F (122A) for determining the representative feature value by weighting one or more feature values extracted from the speech information divided into the female group (F), and the representative And a discriminating unit-F 122B for determining gender and age with reference to the reference DB based on the feature values. As described above, the combination operation unit-F 122A may receive feature values extracted from the sex feature extractor 111, the age-specific feature extractor-FC 113, and the age-specific feature extractor-F 114. .

In addition, the speech operation unit-C 123 may assign a weight to one or more feature values extracted from the speech information divided into the children group C to determine the representative feature value, and the representative operation unit C-123A, and the representative thereof. Determination unit -C (123B) for determining the gender and age with reference to the reference DB based on the feature value. In addition, as described above, the combination operation unit-C 123A may receive feature values extracted from the sex feature extractor 111, the age-specific feature extractor-FC 113, and the age-specific feature extractor-C 115. have.

In order to determine such age and gender, the age and gender may be determined using algorithms such as Gaussian Mixture Model (GMM), Neural Network (NN), and Support Vector Machine (SVM). However, the above-described algorithm is merely an example, and in addition to the above-described algorithm, age and gender can be determined from the feature values using various algorithms.

For example, in the case of using the algorithm of the GMM, each combination calculating unit 121A, 122A, 123A calculates N likelihood values corresponding to the number N of the feature discrimination method or the number N of the plurality of samples, and such N The representative value can be determined from the likelihood values of the dogs. In order to determine the representative value, the combination calculation units 121A, 122A, and 123A may determine the representative value by obtaining an average value of N likelihood values, a maximum value, a minimum value, or summing the total values.

In addition, the combination operation unit 121A, 122A, 123A may determine the representative feature value by assigning a weight value in determining the representative feature value. The weight value may be set according to a situation or a weight value may be set using empirically accumulated information. For example, in a noisy environment, the feature value portion corresponding to the noise band may be set to a low weight, and the weight value may be assigned to a feature value portion corresponding to an intermediate band in the general voice band. In addition, each combination operation unit 121A, 122A, 123A may determine the representative feature value by differently weighting each group (male, female, child) by reflecting the voice characteristic.

In the above description, the voice information is divided into a male group, a female group, and a child group. However, when the feature value of the voice information is extracted, and the feature value is ambiguous, the overlapping of the voice information is applied to each group. desirable. That is, in the case of ambiguity of speech information for classifying groups, arithmetic operation is applied to each corresponding group, and finally, through the final discriminating unit 300 in consideration of the similarity, the normal discrimination probability, the reliability, etc. between the results of each discriminating unit. Determine your age and gender.

FIG. 3 is a detailed configuration diagram of the image processor of FIG. 1.

As illustrated in FIG. 3, the image processor 200 includes an image feature extractor 210 extracting a feature value from image information and an image calculator 220 calculating a gender and an age from the extracted feature value. Can be configured.

The image feature extractor 210 may receive image information and extract feature values. The image feature extractor 210 again includes an age-specific feature extractor 211, an age-specific feature extractor-C 212, a sex feature extractor-C 213, a sex feature extractor-A 214, and an age-specific feature. The extractor-M 215 and the age-specific feature extractor-F 216 may be included.

The age-specific feature extracting unit 211 extracts feature values by reflecting age-specific features with respect to the input face information, and divides the input face information into an adult group A or a child group C based on the feature values. . For example, in the case of face information, it is easy to distinguish an adult from a child based on eye size, eye wrinkles, etc., and the age-specific feature extracting unit 211 reflects the age-specific features. Feature values can be extracted.

With respect to the face information divided into the child group C by the age-specific feature extractor 211, the age-specific feature extractor-C 212 extracts feature values by reflecting the age-specific features of the child, and the sex feature extractor- C 213 extracts feature values by reflecting the gender characteristics of the child.

The sex feature extractor-A 214 may extract a feature value by reflecting the sex feature of an adult with respect to face information classified into the adult group A by the feature extractor 211 for each age, and the feature value The face information input based on the data is divided into a male group M and a female group F. FIG.

The age-specific feature extractor-M 215 extracts feature values by reflecting the age-specific features of men with respect to face information classified into the male group M by the gender feature extractor-A 214. In addition, the age-specific feature extractor-F 216 extracts feature values by reflecting the age-specific features of women with respect to face information classified into the female group F by the gender feature extractor-A 214.

The image operator 220 determines gender and age from the image information by using the feature values extracted by the image feature extractor 210.

That is, the image operator 220 refers to the reference DB based on the combinational operator and the representative feature value to determine the representative feature value by reflecting the weight of one or more feature values extracted by the image feature extractor 210, and the gender and age. It may be configured to determine the determining unit.

Also, as illustrated in FIG. 3, the image calculation unit 220 includes a male group M, a female group F, and a children group C grouped by the image feature extracting unit 210 as described above. Each may be configured to have an optimized combination operation unit and discrimination unit. That is, the image calculation unit 220 receives the feature values extracted from the image information divided into the male group, and calculates the age and gender. It may be configured as an image operator-F (222) for receiving age and gender and an image operator-C (223) for calculating age and gender by receiving feature values extracted from voice information divided into child groups.

The image calculating unit-M 221 receives the one or more feature values extracted from the face information divided into the male group M, and determines the representative feature value based on the combination calculating unit-M 222A and the representative feature value. It may include a determination unit -M (222B) for determining the gender and age with reference to the reference DB. Combination operation unit-M (222A) receives the feature value extracted from the face information divided into the male group (M), age-specific feature extractor 211, sex feature extractor-A (214) and age-specific feature extractor The representative feature value may be determined by receiving the feature values extracted by the -M 215.

The image calculating unit-F 222 receives a combination of the feature values extracted from the face information divided into the female group F, and combines the calculating unit-F 223A and the representative feature value and the reference DB. Determination unit -F (223B) for determining the gender and age may be used.

In this case, the combination operation unit -F (223A) targets the face information divided into the female group (F), so the age-specific feature extraction unit 211, sex feature extraction unit -A (214) and age-specific feature extraction unit- Representative feature values may be determined by receiving feature values extracted at F 216.

The image calculating unit-C 223 receives the combination of the feature values extracted from the face information divided into the children group C, and determines the representative calculating unit-C 221A, and the representative feature value and the reference DB. Determination unit -C (221B) for determining the gender and age may be used. In this case, the combination calculation unit-C 221A receives the feature values extracted from the face information divided into the child group C, and thus the age-specific feature extraction unit 211, the age-specific feature extraction unit-C 212, and the sexes. Representative feature values may be determined by receiving feature values extracted by the feature extractor-C 213.

Each determination unit 221B, 222B, and 223B may receive a representative feature value from each combination operation unit 221A, 2 22A, and 223A described above to calculate a gender and an age with reference to the reference DB. Since a detailed description thereof is similar to that described above with reference to the voice operator 120, a detailed description thereof will be omitted.

In addition, even when calculating the age and gender using the image processing unit 200, as described above, the face information is divided into any one group of the male group (M), female group (F) and children group (C). In the case of the following ambiguity, the face information can be applied to each group in duplicate.

Hereinafter, the final determination unit 300 will be described in detail.

The final determining unit 300 receives genders and ages output from some or all of the determining units 121B, 122B, 123B, 221B, 222B, and 223B, and combines the genders and ages to determine the final gender and age. Can be determined.

That is, mutual similarity may be calculated for each of a plurality of input genders and ages, and a gender and age having the highest mutual similarity may be determined as final gender and age. Alternatively, the normal discrimination probability or the reliability index of the plurality of input genders and ages may be identified and stored at each determination, and determined as the final gender and age.

The final determining unit 300 determines the gender and age using the mutual similarity with respect to the gender and ages output from the voice processing unit 100, and also mutually with respect to the gender and ages output from the image processing unit 200. After determining gender and age using similarity, the final gender and age may be determined and output using the two determined gender and age.

Alternatively, the final determination unit 300 may be configured to determine and output the final gender and age by using mutual similarity with respect to all gender and age determination results output from the voice processing unit 100 and the image processing unit 200. have.

Hereinafter, the reference DB will be described in detail.

The reference DB stores gender and age reference feature values or voice and image reference samples. The reference DB may be configured as a relationship model of feature values extracted from face information or voice information and gender and age with respect to the feature values.

Using the feature value-gender and age correspondence relationships stored in the reference DB, the voice operator 120 or the image operator 220 may obtain a gender and an age with reference to the reference feature DB. For example, the determination unit may determine gender and age by using a distance value between the representative feature value and the relationship model of the reference DB.

In addition, the reference DB may include image data and audio data and corresponding gender and age so that gender and age can be directly determined by using video or audio information in a case where it is difficult to extract feature values smoothly.

The image data included in the reference DB may be acquired by, for example, separating a camera and a person by a distance of 0.5 m, 1 m, and 3 m, respectively. At this time, if the separation distance is 3m to shoot all the body of the person included. Such image data may be photographed to be 100 frames for 10 seconds. The detailed DB may be configured by acquiring a face, a head shape, a mustache, a shape of an eyebrow, and the like of each subject using a face detector, a key detector, an eye detector, and the like on the photographed image. The present invention can be practiced to determine feature values using the detailed DB configured in this way.

In the case of voice data included in the reference DB, for example, 50 sentences prepared in advance may be obtained by repeating three times. Such voice data may have various forms such as 16 kHz, 16 bit, and mono type.

In order to have a sample, such a reference DB may be configured with data for 120 people, wherein the gender ratio of males and females is 1: 1 and the ratio for each age group is 1: 1. have.

The reference DB possesses learning ability, and when an operation is performed on gender-age according to an embodiment of the present invention, the resultant value of the operation (representative characteristic value of the operation and the final gender and age) is presently constituted. It is desirable to reconstruct (update) the DB to reflect the data so that the reliability can be continuously improved. Of course, the result value used for updating the database should be a result of which the reliability is confirmed.

4 is a flowchart of a gender-age identification method according to the present invention.

The input unit 10 collects face information and voice information of a specific person to identify gender and age (S100).

From the collected voice information, the voice processor 100 extracts feature values by reflecting age-specific features and gender features, and determines representative feature values with respect to the extracted one or more feature values. The representative feature value is queried from the reference DB to determine gender and age (S200).

In addition, the image processor 200 extracts a feature value by reflecting age-specific features and gender features with respect to face information, and determines a representative feature value with respect to the extracted one or more feature values. The representative feature value is queried in the reference DB to determine gender and age (S300).

The final determination unit 300 finally determines the gender and age in consideration of mutual similarity or probability with respect to at least one gender and age determined by the step S200 and the step S300 (S400).

Hereinafter, referring to FIG. 5, the step of determining gender and age from the voice of FIG. 4 will be described in detail.

In general, it is difficult to distinguish between female voice information and children's voice information, but it is easy to distinguish between female and child voice information and male voice information. The feature values are extracted to classify the male, female and children groups (S210).

As described above, the gender characteristic is first applied to the speech information by using the fact that the gender characteristic difference is large in the speech information, and thus, the calculation can be performed quickly and efficiently.

According to the classification result, the input voice information is classified into a male group or a female and a child group, and at least one characteristic value of each age reflecting an age characteristic of a male is extracted from the speech information classified into a male group (S220).

Also, with respect to the voice information classified into the female and children group, the age-specific feature values reflecting the age-specific characteristics of the female and the child are extracted to distinguish whether the voice information is the female group or the children's group, It distinguishes (S230).

Thereafter, age-specific feature extraction reflecting age-specific features of women is performed with respect to voice information divided into female groups (S240).

In addition, the voice information distinguished by the child group is performed to extract the characteristics of each child's gender and age (S250).

For the feature values extracted in this way, the representative feature value for the voice information is determined, and the gender and age of the subject are determined.

For example, the voice operator 120 determines a representative feature value with respect to one or more feature values extracted by the voice feature extractor 110, and determines gender and age using a reference DB based on the determined representative feature value. The determination of the representative feature value or the determination of the gender and age may be performed for each of the male group, the female group and the child group as described above.

That is, the voice operation-M is performed on the feature values of the voice information divided into male groups (S225), the voice operation-F is performed on the feature values of the voice information divided into female groups (S245), The gender and age may be determined by performing speech operation-C on the feature values of the speech information divided into child groups (S255).

As described above, voice information is first grouped on the basis of easily distinguishable features (eg, voice information is gender-specific), and the characteristics of each group are reflected for each group classified according to the criteria. Using a method of extracting values is one of the great features of the present invention. By using this stepwise extraction method, the present invention can secure the accuracy of identification and can quickly determine the age and gender of the subject by excluding redundancy of operations.

Hereinafter, referring to FIG. 6, the step of determining gender and age based on the image of FIG. 4 will be described in detail.

In the case of video information, it is generally easy to distinguish between adults and children. For example, it is possible to easily distinguish between adult and child by using biometric information such as height or by using a ratio of the size of the face to the size of the eye.

Using this point, in the image similarity determining step of the present invention, a feature value considering the characteristics of each age with respect to the image information (the image information including the face information or the face information. Extraction is performed (S310). Through this step, the input face information can be easily divided into a child group and an adult group.

Thereafter, age-specific features in consideration of age-specific features of the children are extracted with respect to face information divided into children's groups (S320), and gender feature extraction considering the sex features of the children is performed (S330).

With respect to face information divided into adult groups, gender feature extraction is performed in consideration of adult gender characteristics to distinguish face information of an adult group into a male group or a female group (340).

Thereafter, at least one feature value is extracted by using a feature extraction method that takes into account the age-specific characteristics of men with respect to the face information divided by the male group (S350). One or more feature values are extracted (S360).

As described above, the image calculator 220 determines the representative feature value by reflecting the weights of the feature values extracted by the image feature extractor 210, and uses the determined representative feature value and the reference DB to determine the gender and the gender. Determine your age. As shown in FIG. 6, gender and age discrimination based on the image information is preferably performed for each of a child group, a male group, and a female group (S325, S355, and S365).

As mentioned above, although this invention was demonstrated in detail with reference to attached drawing, this is only an illustration, It is clear that various deformation | transformation and a change are possible within the scope of the technical idea of this invention. Therefore, the protection scope of the present invention should not be limited to the above-described embodiment, but should be determined to include the scope according to the description of the following claims and their equivalents.

1 is a block diagram of an embodiment of a gender-age identification apparatus according to the present invention.

FIG. 2 is a detailed configuration diagram of the voice processor of FIG. 1. FIG.

3 is a detailed configuration diagram of the image processor of FIG. 1.

4 is a flowchart of a gender-age identification method according to the present invention.

5 is a detailed flowchart of the voice similarity determining step of FIG. 4.

6 is a detailed flowchart of the image similarity determining step of FIG. 4.

Claims (19)

Collecting image information and audio information; Extracting one or more feature values with respect to the collected voice information, and determining gender and age using voice information for determining gender and age using the extracted feature values; Extracting one or more feature values with respect to the collected image information, and determining gender and age using face information for determining gender and age using the extracted feature values; And And finally determining the gender and age by combining a gender and age determined using the voice information with a gender and age determined using the face information. The final decision on gender and age is Calculating mutual similarity for each of sex and age determined using at least one voice information and gender and age determined using at least one image information; And Determining the gender and age having the highest mutual similarity as the final gender and age. Human gender-age identification method. The method of claim 1, wherein the determining of gender and age using the voice information comprises: A first gender feature extraction step of extracting a feature value by reflecting the gender feature of the voice with respect to the collected input voice information; A first age feature extraction step of extracting a feature value by reflecting the age-specific features of a male with respect to the voice information divided into male groups by the first gender feature extraction step; And A second age feature extraction step of extracting feature values by reflecting age-specific features of women and children with respect to the voice information divided into women and children groups by the first sex feature extraction step; It comprises a gender-age identification method. The method of claim 2, wherein the determining of gender and age using the voice information comprises: A third age-specific feature extraction step of extracting feature values by reflecting the age-specific features of women with respect to the voice information divided into the female group by the second age-specific feature extraction step; And A second sex feature extraction step of extracting a feature value of the voice information divided into child groups by the second age feature extraction step by reflecting a gender feature of a child; It further comprises a sex-age identification method. The method of claim 2 or 3, wherein the feature value, A gender-age identification method in which M samples are extracted by applying different N feature value discrimination methods. The method of claim 1, wherein the determining of gender and age using the voice information comprises: A representative feature value determining step of determining a representative feature value by applying a weight to the extracted one or more feature values; And A discriminating step of determining the gender and age with reference to the reference DB based on the representative feature value; It comprises a gender-age identification method. The method of claim 5, wherein the representative feature value determination step and the determination step, Carried out separately for the male, female and child groups Human gender-age identification method. The method of claim 5, wherein the representative feature value determination step, Determining any one of an average value, a maximum value, a minimum value, and a sum of one or more of the weighted feature values as the representative feature value; It comprises a gender-age identification method. The method of claim 1, wherein the determining of gender and age using the image information comprises: A first feature extraction step of extracting feature values by reflecting features of each age with respect to the collected image information; And A second feature extraction step of classifying adult and child according to the result of the first feature extraction step, and then classifying the group by male, female and child groups to extract one or more feature values for each group; It comprises a gender-age identification method. The method of claim 8, wherein the determining of gender and age using the image information comprises: A representative feature value determining step of determining a representative feature value by applying a weight to the extracted one or more feature values; And And determining the gender and age by using a reference DB based on the representative feature value. The representative feature value determining step and the calculating step are performed for each male group, female group and children group It comprises a gender-age identification method. delete The method of claim 5 or 9, wherein the reference DB, Includes feature values by gender and age, and continuously reconfigures to reflect feature values verified for reliability by gender and age Human gender-age identification method. An input unit for collecting image information and audio information; A voice processor extracting a feature value with respect to the collected voice information and determining a gender and an age from the voice information using the extracted feature value; An image processor extracting a feature value with respect to the collected image information and determining a gender and an age from the image information using the extracted feature value; And And a final determination unit which finally determines the gender and age of a specific person by combining a gender and age determined by the image processor and the gender and age determined by the voice processor, The final determination unit, Calculating mutual similarity for each of at least one age and gender determined by the voice processor or the image processor, and finally determining the gender and age having the highest mutual similarity; Gender-age identification device. The method of claim 12, wherein the voice processing unit, A voice feature extracting unit for extracting feature values by reflecting gender characteristics or age-specific features of the voice with respect to the collected voice information; And A voice operator extracting a representative feature value from the feature value extracted by the voice feature extractor and determining the gender and age using the representative feature value. Gender-age identification device comprising a. The method of claim 13, wherein the voice feature extraction unit, First determining whether the collected voice is a male voice, and then extracting one or more feature values for each male, female, and child group Gender-age identification device. The method of claim 14, wherein the voice calculation unit, Determining representative feature values for each of the male, female and child groups and determining gender and age using the representative feature values Gender-age identification device. The method of claim 12, wherein the image processor, An image feature extracting unit configured to extract a feature value by reflecting gender characteristics or age-specific features of the image with respect to the collected image information; And An image calculator extracts a representative feature value from the feature value extracted by the image feature extractor, and determines the gender and age using the representative feature value. Gender-age identification device comprising a. The method of claim 16, wherein the image feature extraction unit, First determining whether the collected image is an adult or a child, and then extracting one or more feature values for each male, female, and child group Gender-age identification device. The method of claim 17, wherein the image operation unit, Determining representative feature values for each of the male, female and child groups and determining gender and age using the representative feature values Gender-age identification device. delete

Images