KR101293867B1 - Apparatus and Method for Producing Constitution Decision Function Using Sound - Google Patents

Abstract

Disclosed are an apparatus and method for generating a constitution determination function using voice information. According to an embodiment of the present invention, a method of generating a constitution determining function for generating a constitution determining function for determining a constitution of a user may include receiving a plurality of reference voice information including a plurality of voice variables; Extracting at least one main voice variable from the plurality of voice variables; And generating the constitution determination function by using the extracted main speech variable and a preset classification model.

Description

Apparatus and Method for Producing Constitution Decision Function Using Sound}

Embodiments of the present invention relate to an apparatus and method for generating a constitution determining function, and more particularly, to an apparatus and method for generating a function capable of determining the constitution using voice information.

In modern society, interest in health is increasing day by day.

Along with this era of interest, as technologies such as data analysis method and tool by real-time data collection are advanced, it becomes possible to monitor health condition and provide personalized health care service.

In addition, convenience and customization of health services and related systems are being strengthened due to diversification of customers 'demands and improvement of expectations according to changes in consumers' consciousness. Based on accumulated personal health data, prevention of lifestyle- Personalized health care projects such as weight management are rapidly growing.

If the past health care services were centered on healthcare, which is limited to the treatment of illnesses in hospitals and medical institutions, patients' health care needs in recent years have been limited to the prevention of disease And health maintenance.

In other words, as the level of living improves and the interest in quality of life and well-being / wellness grows, consumers are gradually increasing their preference for proactive health management such as health status measurement and proper exercise management.

According to an embodiment of the present invention, a method of generating a constitution determining function for generating a constitution determining function for determining a constitution of a user may include receiving a plurality of reference voice information including a plurality of voice variables; Extracting at least one main voice variable from the plurality of voice variables; And generating the constitution determination function by using the extracted main speech variable and a preset classification model.

According to an embodiment of the present invention, a method of generating a constitution determining function for generating a constitution determining function for determining a constitution of a user may include receiving a plurality of reference voice information including a plurality of voice variables; Classifying the plurality of reference voice information by gender and age group; Extracting at least one main voice variable by the gender and age group from the plurality of voice variables; And generating the constitution determination function for each gender and age group by using the main voice variable extracted for each gender and age group, reference voice information classified for each gender and age group, and a preset classification model.

An apparatus for generating a constitution determining function, which generates a constitution determining function for determining a constitution of a user according to an embodiment of the present invention, includes: an input unit configured to receive a plurality of reference voice information including a plurality of voice variables; An extraction unit for extracting at least one main voice variable from the plurality of voice variables; And a generator configured to generate the constitution determination function by using the extracted main voice variable and a preset classification model.

An apparatus for generating a constitution determining function, which generates a constitution determining function for determining a constitution of a user according to an embodiment of the present invention, includes: an input unit configured to receive a plurality of reference voice information including a plurality of voice variables; A processing unit classifying the plurality of reference voice information by gender and age group; An extraction unit for extracting at least one main voice variable from the plurality of voice variables by the gender and age group; And a generation unit configured to generate the constitution determination function for each gender and age group by using the main voice variable extracted for each gender and age group, reference voice information classified for each gender and age group, and a preset classification model.

According to embodiments of the present invention, a constitution determining function that may be used to determine the constitution of a user using voice may be generated.

1 is a diagram illustrating an overall algorithm of a method of determining a user's constitution using voice information according to an embodiment of the present invention.
2 is a diagram illustrating a configuration of an apparatus for generating a constitution determination function according to an embodiment of the present invention.
3 and 4 are diagrams showing the difference between voice variable values according to gender and age group.
5 is a diagram illustrating a configuration of an apparatus for generating a constitution determination function for generating a constitution determination function for each gender and age group according to an embodiment of the present invention.
6 is a flowchart illustrating a method of generating a constitution determination function according to an embodiment of the present invention.
7 is a flowchart illustrating a method of generating a constitution determination function for generating a constitution determination function for each gender and age group according to an embodiment of the present invention.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to or limited by the embodiments. Like reference symbols in the drawings denote like elements.

1 is a diagram illustrating an overall algorithm of a method of determining a user's constitution using voice information according to an embodiment of the present invention.

Referring to FIG. 1, a method of determining a user's constitution 130 using voice information 100 according to an embodiment of the present invention includes voice information 100 for a user and personal information about a user's image ( 110) can be input.

The personal information 110 may include information about the user's personal information such as age, gender, height, weight, BMI (body mass index), occupation, and education level of the user.

The voice information 100 represents information about the voice of the user. At this time, the voice information 100 may include a value sensed for a plurality of voice variables associated with the voice of the user. According to an embodiment, the voice information may include a voice variable for the vowel voice 101 of the user and a voice variable for the sentence voice 102.

The input image information 110 and the voice information 100 may be input as variables of the constitution determining function 120 for determining the constitution 130 of the user.

The constitution determination function 120 is a function for determining the constitution of the user 130 and may be a function capable of calculating the constitution of the user by receiving the voice information 100 and the personal information 110 of the user. have. The constitution determination function 120 may be generated using a plurality of clinical data and reference voice data. For example, a plurality of clinical data and reference voice data may be input to a classification model to calculate a function, and the calculated function may be used as the constitution determination function 120. The classification model may be linear discriminant analysis (LDA), logistic regression analysis, neural network analysis, support vector machine analysis, decision tree analysis, or the like.

In addition, the user's constitution 130 may be determined by inputting the received voice information 100 and the personal information 110 to the constitution determining function 120. That is, the user's constitution 130 may be determined as one of the sun 131, the lunar 132, the yangyang 133, and the noise 134.

As described above, when the voice information 100 and the personal information 110 for the user are provided, the constitution determination function 120 may be easily determined by inputting the constitution determination function 120. ) Plays a major role, and therefore, the present specification proposes a new method for generating the constitution determination function 120.

2 is a diagram illustrating a configuration of an apparatus for generating a constitution determination function according to an embodiment of the present invention.

Referring to FIG. 2, the constitution determination function generating device 200 according to an embodiment of the present invention may include an input unit 210, an extractor 220, and a generator 230.

The constitution determination function generating device 200 represents an apparatus for generating a constitution determination function for determining the constitution of a user.

The input unit 210 may receive a plurality of reference voice information 201 for each voice of a plurality of clinical trial subjects. In this case, the reference voice information 201 may include a plurality of voice variables 202. The negative variable 202 will be described in detail later.

As described above, the input unit 210 collects voice information about a plurality of clinical trial subjects, so that the constitution determination function generating apparatus 200 may collect statistical data for generating a constitution determination function, and the collected data Based on the constitution determination function can be generated.

The extractor 220 may extract at least one main voice variable from the plurality of voice variables 202.

According to one side of the present invention, the extractor 220 extracts the main voice variable from the plurality of voice variables 202 using the values of the coefficient of variation (CV) for the plurality of voice variables 202. Can be. In detail, the extractor 220 may calculate values of coefficients of variation (CV) for the plurality of voice variables 202. In addition, the extraction unit 220 may perform an operation of reducing the voice dimension of the plurality of voice variables 202 based on the calculated value of the variation coefficient. In this case, the extractor 220 may extract the main voice variable from the voice variable 202 whose voice dimension is reduced.

According to another aspect of the present invention, the extraction unit 220 may extract the main voice variable by performing a correlation between the variables for the plurality of voice variables 202 to minimize. For example, the extractor 220 may calculate values of a variation inflation factor (VIF) for the plurality of voice variables 202. In addition, the extractor 220 may extract the main voice variable by removing the voice variable 202 having a calculated VIF value greater than or equal to the reference value among the plurality of voice variables 202. According to another exemplary embodiment, the extractor 220 may apply an independent variable selection method such as principal component analysis instead of the value of VIF to minimize correlation between variables.

According to another aspect of the present invention, the extraction unit 220 may extract the main voice variable from the plurality of voice variables 202 using the value of the InterQuartile Range (IQR) for the plurality of voice variables 202. . In detail, the extractor 220 may calculate IQR values for the plurality of voice variables 202. In addition, the extractor 220 may extract the main voice variable by removing the voice variable 202 corresponding to the outlier from the plurality of voice variables 202 using the value of IQR.

The generation unit 230 may generate a constitution determination function by using the extracted main voice variable and a preset classification model. According to an embodiment, the classification model may include at least one of logistic regression analysis, neural network analysis, support vector machine analysis, decision tree analysis, and linear discriminant analysis (LDA). It may include.

According to an aspect of the present invention, the generation unit 230 may calculate a correlation between the plurality of voice variables and the constitution using the reference voice information 201 and the classification model. The correlation calculated at this time may be utilized as a coefficient for each voice variable of the classification model. For example, the generation unit 230 calculates a correlation between the plurality of voice variables and the constitutions (sun, lunar, literacy, noise), and uses the values for the calculated correlations as coefficients for each variable of the classification model. It can be utilized. Therefore, when coefficients for each voice variable of the classification model are calculated, the generation unit 230 may input the calculated coefficient and the main voice variable to the classification model to generate a constitution determination function.

Hereinafter, the voice information and the voice variable will be described in detail.

The voice information may include a plurality of voice variables associated with the voice of the user or clinical trial subject. Further, the negative variable may include a vowel variable associated with vowel speech of a user or a clinical trial subject and a negative variable associated with sentence utterance of a user or a clinical trial subject. At this time, the vowel utterance refers to a voice recorded by the user uttering a predetermined vowel (a, a, e, i, i, o, u, etc.), and the sentence utterance is the user. Indicates a voice that reads a preset sentence.

According to one embodiment, the vowel variable is a JITA representing the average frequency of the pitch, the formant, the frequency difference, the frequency ratio, the average rate of change of the pitch period, the percentage of the JITA, and the RAP associated with the JITA. And PPQ associated with JITA.

Tables 1 and 2 summarize the vowel variables. Table 1 shows the basic variables for vowel vowels and Table 2 shows additional variables for vowel vowels. In this case, Tables 1 and 2 are only examples, and the present invention is not limited to the contents of Tables 1 and 2.

Table 1

Table 2

At this time, in Table 1 and Table 2, F0 represents the average frequency of the pitch. F0 may be calculated as an average of pulse repetition rates per second caused by tremors of the vocal cords.

Equation 1

In Table 1 and Table 2, aF0, eF0, iF0, oF0, and uF0 each represent the average frequency of the odor for each of a (a), (e), (i), (o), and (u) vocalizations. Indicates.

Also, in Tables 1 and 2, F1, F2, F3, and F4 represent formants, which are resonance frequencies occurring in the tremor and voice structure of the vocal cords. If a, e, i, o, u is preceded by F1, F2, F3, F4, this indicates the formant for each vocalization of the attached vowel. For example, aF1 represents a formant for a (a) vocalization. F1, which represents the first formant frequency, may be related to whether the jaw and lip is opened or closed, and F2, which represents the second formant frequency, may be related to the shape of the tongue body. F3, which represents the formant frequency, may be related to the position of the tip of a Tongue, and F4, which represents the fourth formant frequency, may vary according to various passages of the vocal organs.

Frequency difference variables (aF0-eF0, aF0-iF0, etc.) in Table 2 are skxksose relative frequency differences. For example, aF0-eF0 represents the difference between the frequencies of a (a) utterance and (e) utterances, and aF0-iF0 represents the frequencies of a (a) utterances and frequencies of two (i) utterances. Indicates a difference.

The frequency ratio variables (aF0 / aF1, aF0 / F2, etc.) of Table 1 represent the relative magnitude ratios between frequencies.

JITA in Table 1 shows the average rate of change of the pitch period. JITA is a speech variable that represents the rate of change of the speaker's pitch period, and expresses irregularities in the period. JITA according to an embodiment may be calculated as Equation 2.

Equation 2

JITT in Table 1 represents negative variables representing JITA as a percentage.

The RAP in Table 1 represents a parameter that smoothes the value around JITA. The concept of JITA can reduce errors by increasing the interval. JITA can be calculated as shown in Equation 3. The range of values around JITA can be set by the user.

Equation 3

PPQ in Table 1 represents a parameter that smoothes the value around JITA. The concept of JITA can reduce errors by increasing the interval. JITA can be calculated like Equation 4. The range of values around JITA can be set by the user.

Equation 4

A sentence variable according to an embodiment may include a sentence mean fundamental frequency, a standard deviation of a sentence mean fundamental frequency, a tenth percentile value of a sentence mean fundamental frequency, a 50th percentile value of a sentence mean fundamental frequency, and a 90th percentile of a sentence mean fundamental frequency. Value, PHL, sentence strength, standard deviation of sentence strength, 10th percentile value of sentence strength, 50th percentile value of sentence strength, 90th percentile value of sentence strength, IHL, Pearson's correlation coefficient between mean fundamental frequency and intensity signal magnitude per interval And sentence speed.

Tables 3 and 4 summarize the negative variables (variables) for sentence utterances, Table 3 shows the basic variables for sentence utterances, and Table 4 shows additional variables for sentence utterances. At this time, Tables 3 and 4 are only examples, and the present invention is not limited to the contents of Tables 3 and 4.

TABLE 3

Table 4

In Tables 3 and 4, SF0 represents the sentence average fundamental frequency. According to an embodiment, SF0 may be a sentence average fundamental frequency based on the section 4 ms.

SSTD in Table 3 represents the standard deviation of the sentence mean fundamental frequency.

P10 in Table 3 represents the 10th percentile value of the average fundamental frequency, P50 represents the 50th percentile value of the average fundamental frequency, and P90 represents the 90th percentile value of the average fundamental frequency.

PHL of Table 3 shows the value of (P90-P50) / (P50-P10).

SITS in Table 3 represents the sentence strength and may be calculated as shown in Equation 5.

Equation 5

The SISTD in Table 3 represents the standard deviation of sentence strength.

I10 in Table 3 represents the tenth percentile value of intensity, I50 represents the 50th percentile value of intensity, I90 represents the 90th percentile value of intensity, and IHL represents the value of (I90-I50) / (I50-I10). .

CORR in Table 3 shows Pearson's correlation coefficient between interval F0 and strength signal magnitude.

SSPD of Table 3 represents a sentence speed, which is a sentence read time, and may be calculated as shown in Equation 6.

Equation 6

The negative variable may also include features for the combination of vowel and sentence vowels. According to an embodiment, the phonetic variable may include the harmonics between vowel and sentence utterances and the love rate between the overtones.

The overtones between vowel speech and sentence speech can represent powers of 120-240 Hz, 240-480 Hz, 480-960 Hz, 960-1920 Hz, and 1920-3840 Hz.

3 and 4 are diagrams showing the difference between voice variable values according to gender and age group.

Referring to FIG. 3, the average frequencies (aF0, aF1) (310, 320) of the vowels (a) and the MFCC (aMFCC1) 330 of the negative (a) sound are different according to the genders of men and women. May appear.

In addition, referring to FIG. 4, when the graph of time-varying 410 of the average frequency (aF0) 401 of the vowels (a) is shown, the male is aFO up to 70's after the metamorphic period (after 20's). The value 401 appears to increase continuously. However, in the graph of time-dependent change of the average frequency (aF0) (401) for the vowels (420) for the excitation, the excitation rate of the aFO (401) continues to decrease linearly from the teens to the seventies. appear.

Therefore, as shown in FIG. 3 and FIG. 4, the difference in the value for the negative variable may vary greatly according to gender and age group. Therefore, in order to prevent the reliability deterioration due to the above-described difference, the constitution determination function may be generated by classifying by gender and age group. Hereinafter, an operation of generating the constitution determination function by classifying the constitution determination function by gender and age group will be described in detail.

5 is a diagram illustrating a configuration of an apparatus for generating a constitution determination function for generating a constitution determination function for each gender and age group according to an embodiment of the present invention.

Referring to FIG. 5, an apparatus 500 for generating a constitution determination function for each gender and age group according to an embodiment of the present invention includes an input unit 510, a processor 520, an extractor 530, and a generator. It may include a portion 540.

The input unit 510 may receive a plurality of reference voice information 501 for the voices of each of the plurality of clinical trial subjects. In this case, the reference voice information 501 may include a plurality of voice variables 502.

The processor 520 may classify the received plurality of reference voice information 501 by gender and age group. For example, the processor 520 may classify the plurality of reference voice information 501 into a male and a female. In addition, the processor 520 may classify the plurality of reference voice information 501 into 20s / 30-40s / 50s or more. In addition, the processor 520 may classify the plurality of pieces of reference voice information 501 in consideration of gender and age group.

The extractor 530 may extract at least one main voice variable from the plurality of voice variables 502 by gender and age group. Therefore, according to the exemplary embodiment, the main negative variable may be extracted differently according to gender and age group. The extractor 530 extracts the main voice variable by gender and age group, and the same method of extracting the main voice variable described with reference to FIG. 2 may be applied.

The generation unit 540 uses the (1) the main voice variable extracted by gender and age group, (2) the reference voice information 501 classified by gender and age group, and (3) a pre-set classification model to determine a constitution determination function. Can be created by gender and age group.

According to an embodiment, the classification model may include at least one of logistic regression analysis, neural network analysis, support vector machine analysis, decision tree analysis, and linear discriminant analysis (LDA). It may include.

According to an aspect of the present invention, the generation unit 540 may calculate a correlation between the plurality of voice variables 502 and the constitution using the reference voice information 501 classified by gender and age group, and a classification model. . The correlation calculated at this time may be utilized as a coefficient for each voice variable 502 of the classification model for each gender and age group. For example, the generation unit 540 calculates a correlation between the plurality of voice variables 502 and the constitutions (sun, lunar, literacy, noise), and calculates values for the calculated correlations by gender and age group. Can be used as a factor for each variable in. Therefore, when coefficients for each voice variable 502 of the classification model are calculated, the generation unit 540 inputs the coefficients and main voice variables calculated in the classification model for each gender and age group, and calculates a constitution determination function for each gender and age group. Can be generated.

Table 5 is a table showing the results of the calculation unit 540 calculated the coefficients of the classification function of each constitution for the voice variable 502 classified into 30-40 male. Table 5 is only one embodiment, the present invention is not limited thereto.

Table 5

In addition, Table 6 is a table showing the result of the generation unit 540 calculated the coefficient of the classification function of each constitution for the voice variable 502 classified into 30-40 women. Table 6 is only one embodiment, but the present invention is not limited thereto.

Table 6

Therefore, the generator 540 calculates the coefficients for each of the main voice variables of the classification model for each gender and age group, as shown in Tables 5 and 6, and then inputs the calculated coefficients and the main voice variables to the classification model. A constitution judgment function can be generated for each age group.

6 is a flowchart illustrating a method of generating a constitution determination function according to an embodiment of the present invention.

Referring to FIG. 6, the method of generating a constitution determination function according to an embodiment of the present invention may generate a constitution determination function for determining a constitution of a user.

The method of generating a constitution determination function may receive a plurality of reference voice information for each of a plurality of clinical test subjects' voices (610). In this case, the reference voice information may include a plurality of voice variables. As described above, the method of generating a constitution determination function may collect statistical data for generating a constitution determination function by collecting voice information of a plurality of clinical trial subjects, and generate a constitution determination function based on the collected data. can do.

The method of generating a constitution determination function may extract at least one main voice variable from the plurality of voice variables (620).

According to one aspect of the present invention, the method for generating a constitution determination function may extract a main voice variable from a plurality of voice variables by using values of a coefficient of variation (CV) for a plurality of voice variables. In detail, the method of generating a constitution determination function may calculate values of coefficients of variation (CV) for a plurality of voice variables. In addition, the method of generating a constitution determination function may perform an operation of reducing the speech dimension of the plurality of speech variables based on the calculated variation coefficient. In this case, the method of generating a constitution determination function may extract a main voice variable from the voice variable whose voice dimension is reduced.

According to another aspect of the present invention, the method for generating a constitution determination function may extract a main voice variable by performing a correlation between variables for a plurality of voice variables. For example, the method of generating a constitution determination function may calculate values of a variation inflation factor (VIF) for a plurality of voice variables. In addition, the method of generating a constitution determination function may extract a main voice variable by removing a voice variable having a calculated VIF value equal to or greater than a reference value among a plurality of voice variables. According to another embodiment, the method of generating a constitution determination function may perform an operation of minimizing correlation between variables by applying an independent variable selection method such as principal component analysis instead of the value of VIF.

According to another aspect of the present invention, the method for generating a constitution determination function may extract a main voice variable from a plurality of voice variables using the value of an interquart range (IQR) for a plurality of voice variables. In detail, the method of generating a constitution determination function may calculate values of IQRs for a plurality of voice variables. In addition, the method of generating a constitution determination function may extract a main voice variable by removing a voice variable corresponding to an outlier from a plurality of voice variables using the value of IQR.

The method of generating a constitution determination function may generate a constitution determination function by using the extracted main speech variable and a preset classification model (630). According to an embodiment, the classification model may include at least one of logistic regression analysis, neural network analysis, support vector machine analysis, decision tree analysis, and linear discriminant analysis (LDA). It may include.

According to one aspect of the present invention, the method of generating a constitution determination function may calculate a correlation between a plurality of speech variables and a constitution using reference speech information and a classification model. The correlation calculated at this time may be utilized as a coefficient for each voice variable of the classification model. For example, the method of generating a constitution determination function calculates a correlation between a plurality of voice variables and a constitution (sun, lunar, literacy, noise), and calculates the value of the calculated correlation as a coefficient for each variable of the classification model. It can be utilized. Therefore, when the coefficient for each voice variable of the classification model is calculated, the constitution determination function generating method may generate the constitution determination function by inputting the calculated coefficient and the main voice variable to the classification model.

The voice information may include a plurality of voice variables associated with the voice of the user or clinical trial subject. Further, the negative variable may include a vowel variable associated with vowel speech of a user or a clinical trial subject and a negative variable associated with sentence utterance of a user or a clinical trial subject. At this time, the vowel utterance refers to a voice recorded by the user uttering a predetermined vowel (a, a, e, i, i, o, u, etc.), and the sentence utterance is the user. Indicates a voice that reads a preset sentence.

According to one embodiment, the vowel variable is a JITA representing the average frequency of the pitch, the formant, the frequency difference, the frequency ratio, the average rate of change of the pitch period, the percentage of the JITA, and the RAP associated with the JITA. And PPQ associated with JITA.

A sentence variable according to an embodiment may include a sentence mean fundamental frequency, a standard deviation of a sentence mean fundamental frequency, a tenth percentile value of a sentence mean fundamental frequency, a 50th percentile value of a sentence mean fundamental frequency, and a 90th percentile of a sentence mean fundamental frequency. Value, PHL, sentence strength, standard deviation of sentence strength, 10th percentile value of sentence strength, 50th percentile value of sentence strength, 90th percentile value of sentence strength, IHL, Pearson's correlation coefficient between mean fundamental frequency and intensity signal magnitude per interval And sentence speed.

The negative variable may also include features for the combination of vowel and sentence vowels. According to an embodiment, the phonetic variable may include the harmonics between vowel and sentence utterances and the love rate between the overtones.

The overtones between vowel speech and sentence speech can represent powers of 120-240 Hz, 240-480 Hz, 480-960 Hz, 960-1920 Hz, and 1920-3840 Hz.

7 is a flowchart illustrating a method of generating a constitution determination function for generating a constitution determination function for each gender and age group according to an embodiment of the present invention.

Referring to FIG. 7, in the method of generating a constitution determination function for generating a constitution determination function for each gender and age group according to an embodiment of the present invention, a plurality of reference voice information of each voice of a plurality of clinical subjects may be received ( 710). In this case, the reference voice information may include a plurality of voice variables.

The method of generating a constitution determination function may classify the plurality of input reference voice information by gender and age group (720). For example, the method of generating a constitution determination function may classify a plurality of reference voice information into a male and a female. In addition, the method of generating a constitution determination function may classify a plurality of reference voice information into 20s / 30-40s / 50s or more. In addition, the method of generating a constitution determination function may classify a plurality of reference voice information in consideration of gender and age group.

The method of generating a constitution determination function may extract at least one main voice variable from a plurality of voice variables by gender and age group (730). Therefore, according to the exemplary embodiment, the main negative variable may be extracted differently according to gender and age group. The method of extracting the main voice variable described with reference to FIG. 2 may be applied to an operation of extracting the main voice variable by gender and age group.

The method of generating a constitution judgment function is performed by using (1) main voice variables extracted by gender and age group, (2) reference voice information classified by gender and age group, and (3) preset classification model. The generation may be performed separately (740).

According to an embodiment, the classification model may include at least one of logistic regression analysis, neural network analysis, support vector machine analysis, decision tree analysis, and linear discriminant analysis (LDA). It may include.

According to an aspect of the present invention, the method of generating a constitution determination function may calculate a correlation between a plurality of voice variables and a constitution by using reference voice information and a classification model classified by gender and age group. The correlation calculated at this time may be utilized as a coefficient for each voice variable of the classification model for each gender and age group. For example, the method of generating a constitution determination function calculates a correlation between a plurality of voice variables and constitutions (sun, lunar, literacy, noise), and calculates the values of the calculated correlations for each variable of the classification model by gender and age group. It can be used as a coefficient for. Therefore, when the coefficient for each voice variable of the classification model is calculated, the constitution determination method generating method may generate the constitution determination function for each gender and age group by inputting the coefficient and the main voice variable calculated in the classification model for each gender and age group. have.

Embodiments according to the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as a hard disk, a floppy disk, and a magnetic tape; optical media such as CD-ROM and DVD; magnetic recording media such as a floppy disk; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the equivalents of the claims, as well as the claims.

200: constitution determination function generator
201: Reference voice information
202: negative variable
210:
220: extraction unit
230: generation unit

Claims (12)

delete In the method for generating a constitution determination function by the apparatus for generating a constitution determination function for determining the constitution of the user, the method,
Receiving a plurality of input voice information including a plurality of voice variables by an input unit of the device;
Extracting, by the extracting unit of the device, at least one main speech variable from the plurality of speech variables; And
Generating, by the generation unit of the device, the constitution determination function by using the extracted main speech variable and a preset classification model;
Lt; / RTI >
Generating the constitution determination function
Calculating coefficients of the classification model for the extracted main speech variables using the plurality of speech variables,
And generating the constitution determination function by inputting the calculated coefficient and the main speech variable to the classification model. delete delete In the method for generating a constitution determination function by the apparatus for generating a constitution determination function for determining the constitution of the user, the method,
Receiving a plurality of input voice information including a plurality of voice variables by an input unit of the device;
Extracting, by the extracting unit of the device, at least one main speech variable from the plurality of speech variables; And
Generating, by the generation unit of the device, the constitution determination function by using the extracted main speech variable and a preset classification model;
Lt; / RTI >
Extracting the main negative variable
Calculating a value of a coefficient of variation (CV) for the plurality of voice variables;
Reducing a speech dimension for the plurality of speech variables based on the calculated variation coefficients; And
Extracting the main speech variable from the speech variable whose speech dimension is reduced
Constitution determination function generation method comprising a. In the method for generating a constitution determination function by the apparatus for generating a constitution determination function for determining the constitution of the user, the method,
Receiving a plurality of input voice information including a plurality of voice variables by an input unit of the device;
Extracting, by the extracting unit of the device, at least one main speech variable from the plurality of speech variables; And
Generating, by the generation unit of the device, the constitution determination function by using the extracted main speech variable and a preset classification model;
Lt; / RTI >
Extracting the main negative variable
And generating the main speech variable by performing correlation minimization between the plurality of speech variables. The method according to claim 6,
The operation of minimizing the correlation between the variables
Calculating a value of a variation inflation factor (VIF) for the plurality of voice variables;
And extracting the main voice variable by removing a voice variable having a value of the calculated VIF equal to or greater than a reference value among the plurality of voice variables. In the method for generating a constitution determination function by the apparatus for generating a constitution determination function for determining the constitution of the user, the method,
Receiving a plurality of input voice information including a plurality of voice variables by an input unit of the device;
Extracting, by the extracting unit of the device, at least one main speech variable from the plurality of speech variables; And
Generating, by the generation unit of the device, the constitution determination function by using the extracted main speech variable and a preset classification model;
Lt; / RTI >
Extracting the main negative variable
Calculating values of InterQuartile Range (IQR) for the plurality of speech variables;
And extracting the main voice variable by removing a voice variable corresponding to an outlier from the plurality of voice variables using the value of the IQR. delete A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 2, 5, 6, 7, or 8. delete delete

Images