Voice feature amount calculation method, voice feature amount calculation device, and oral function evaluation device

Abstract

The sound feature amount calculation method, which is executed by a computer, calculates the feature amount of the sound of an evaluated person (U) from the sound spoken by the evaluated person (U), and acquires sound data obtained by collecting the sound spoken by the evaluated person (U); a sound pressure adjustment step for adjusting the sound pressure of the sound data on the basis of a first average intensity of sound collected during a period in which the evaluated person (U) does not emit sound in the acquired sound data; a feature quantity including at least a feature quantity related to the sound pressure is calculated from the sound data after the sound pressure is adjusted.

Description

Technical Field

[0001] The present invention relates to a voice feature quantity calculation method for calculating the voice feature quantity of an evaluated person, a voice feature quantity calculation device, and an oral function evaluation device using the voice feature quantity calculation device. Background Art

[0002] A method is disclosed in which a device for evaluating the ingestion and swallowing function is worn on the neck of an evaluation subject, and a pharyngeal movement feature is acquired as an ingestion and swallowing function evaluation marker to evaluate the ingestion and swallowing function of the evaluation subject (for example, see Patent Document 1).

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Publication No. 2017-23676 Summary of the invention

[0006] Problems to be solved by the invention

[0007] However, in the method disclosed in the above-mentioned patent document 1, in order to evaluate oral functions such as swallowing function, the person being evaluated needs to wear an instrument, which sometimes causes discomfort to the person being evaluated and creates a burden on the person being evaluated. In addition, although oral function can also be evaluated by visual examination, questioning or palpation by experts such as dentists, dental hygienists, speech and hearing therapists or internists, the elderly, although they have been choking or spilling food due to the influence of aging, ignore the reduction in oral function as a natural symptom of the elderly. Since the reduction in oral function is ignored, for example, malnutrition due to the reduction in food intake will occur, and malnutrition will lead to reduced immunity. In addition, it is easy to swallow by mistake, and the result of swallowing by mistake and reduced immunity may lead to pneumonia caused by swallowing by mistake, creating a vicious circle.

[0008] Even without using such a method, the oral function of the person being evaluated can be evaluated based on the sound spoken by the person being evaluated. However, there is a problem in the accuracy of calculation of the feature amount of the sound used for the evaluation.

[0009] Therefore, an object of the present invention is to provide a voice feature quantity calculation method and the like capable of more appropriately calculating a voice feature quantity based on a voice of an evaluation subject.

[0010] Means for solving problems

[0011] A sound feature quantity calculation method related to a technical solution of the present invention is a sound feature quantity calculation method executed by a computer, and is used to calculate the feature quantity of the sound of the above-mentioned person being evaluated based on the sound spoken by the above-mentioned person being evaluated. The method obtains sound data obtained by collecting the sound spoken by the above-mentioned person being evaluated; based on the first average intensity of the sound collected during the period when the above-mentioned person being evaluated did not make any sound in the obtained sound data, the sound pressure of the above-mentioned sound data is adjusted; based on the above-mentioned sound data after the sound pressure is adjusted, the above-mentioned feature quantity at least including the feature quantity related to the sound pressure is calculated.

[0012] In addition, a sound feature quantity calculation device related to a technical solution of the present invention is a sound feature quantity calculation device that calculates the feature quantity of the sound of the above-mentioned person being evaluated based on the sound spoken by the above-mentioned person being evaluated, and comprises: an acquisition unit, which acquires sound data obtained by collecting the sound spoken by the above-mentioned person being evaluated; a sound pressure adjustment unit, which adjusts the sound pressure of the above-mentioned sound data based on the first average intensity of the sound collected during the period when the above-mentioned person being evaluated did not make any sound in the acquired above-mentioned sound data; and an extraction unit, which calculates the above-mentioned feature quantity by extracting the above-mentioned feature quantity including at least the feature quantity related to the sound pressure from the above-mentioned sound data after the sound pressure is adjusted.

[0013] In addition, an oral function evaluation device according to a technical solution of the present invention comprises: the sound feature quantity calculation device described above; a calculation unit, which calculates an estimated value of the oral function of the above-mentioned person being evaluated based on an estimation formula of the feature quantity related to the above-mentioned sound pressure among the above-mentioned feature quantity extracted from the above-mentioned sound data, and the above-mentioned feature quantity extracted from the above-mentioned sound data after adjusting the sound pressure; and an evaluation unit, which uses an oral function evaluation index to determine the calculated above-mentioned estimation value, thereby evaluating the reduced state of the oral function of the above-mentioned person being evaluated.

[0014] Effects of the Invention

[0015] According to the oral function evaluation method and the like of the present invention, it is possible to more appropriately calculate the feature amount of the voice from the voice of the person being evaluated. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a diagram showing the structure of an oral function evaluation system according to an embodiment.

[0017] Figure 2 This is a block diagram showing a characteristic functional structure of the oral function evaluation system according to the embodiment.

[0018] Figure 3A This is a flowchart showing a processing procedure for evaluating the oral function of an evaluation subject based on the oral function evaluation method according to the embodiment.

[0019] Figure 3BThis is a flowchart showing a processing procedure for sound data used in the oral function evaluation method according to the embodiment.

[0020] Figure 3C This is a diagram showing an example of information output in the oral function evaluation method according to the embodiment.

[0021] Figure 3D This is a flowchart showing a procedure for processing voice data used in an oral function evaluation method according to another example of the embodiment.

[0022] Figure 3E This is a first diagram for explaining adjustment of the sound pressure of audio data according to the embodiment.

[0023] Figure 3F FIG. 2 is a second diagram for explaining adjustment of the sound pressure of audio data according to the embodiment.

[0024] Figure 3G FIG. 3 is a diagram for explaining adjustment of the sound pressure of audio data according to the embodiment.

[0025] Figure 3H This is a graph showing the relationship between the processing of adjusting the sound pressure and the accuracy (estimation accuracy) in the oral function evaluation method according to the embodiment.

[0026] Figure 4 This is a diagram showing an overview of a method for acquiring the voice of an evaluation subject in the oral function evaluation method according to the embodiment.

[0027] Figure 5A This is a diagram showing an example of voice data in which the evaluee speaks "えをかくことにきめたよ".

[0028] Figure 5B This is a graph showing an example of a change in the formant frequency of the voice of the person being evaluated saying "えをかくことにきめたよ".

[0029] Figure 6 This is a diagram showing an example of audio data in which the subject of evaluation repeatedly utters “からからから…”.

[0030] Figure 7 1 is a diagram showing an example of audio data of the person being evaluated saying “いったい”.

[0031] Figure 8 This is a diagram showing an example of Japanese syllables or fixed sentence patterns, and Chinese syllables or fixed sentence patterns that are similar to the degree of tongue movement or mouth opening and closing when pronouncing them.

[0032] Fig. 9A It is a diagram showing the International Phonetic Alphabet of vowels.

[0033] Fig. 9B It is a diagram showing the International Phonetic Alphabet of consonants.

[0034] Fig. 10A 2 is a diagram showing an example of audio data showing the audio of the person being evaluated saying “gao dao wu da ka ji ke da yi wu zhe”.

[0035] Fig. 10B : is a graph showing an example of changes in the formant frequency of the sound of the evaluated person uttering “gao dao wu da ka ji ke da yi wu zhe”.

[0036] Fig.11 This is a diagram showing an example of oral function evaluation index.

[0037] Fig.12 This is a diagram showing an example of evaluation results for each element of oral function.

[0038] Fig.13 This is a diagram showing an example of evaluation results for each element of oral function.

[0039] Fig.14 This is an example of preset data used when providing advice on oral function. DETAILED DESCRIPTION

[0040] Hereinafter, the embodiments are described with reference to the accompanying drawings. In addition, the embodiments described below all represent inclusive or specific examples. The numerical values, shapes, materials, constituent elements, configuration positions of constituent elements and connection forms, steps, the order of steps, etc. shown in the following embodiments are examples and are not intended to limit the present invention. In addition, regarding the constituent elements of the following embodiments that are not recorded in the independent claims representing the highest concept, they are described as arbitrary constituent elements.

[0041] In addition, each figure is a schematic diagram, and does not necessarily illustrate the exact figure. In addition, in each figure, the same reference numeral is given to substantially the same structure, and the repeated description may be omitted or simplified.

[0042] (Implementation Method)

[0043] [Elements of oral function]

[0044] The present invention relates to a method for evaluating a decrease in oral function, etc. Oral function includes various elements.

[0045] For example, the elements of oral function include tongue coating adhesion, oral mucosal moistness, bite force, tongue pressure, cheek pressure, number of remaining teeth, swallowing function, and chewing function. Here, tongue coating adhesion, oral mucosal moistness, bite force, tongue pressure, and chewing function are briefly described.

[0046] The degree of adhesion of the tongue coating indicates the extent of bacterial or food deposition on the tongue. In the case of no tongue coating or thin tongue coating, it can be known that there is a mechanical friction environment (food intake, etc.), there is a cleaning effect of saliva, and the swallowing movement (tongue movement) is normal. On the other hand, in the case of thick tongue coating, it can be known that because the tongue movement is poor and it is difficult to take in food, it may lead to malnutrition or insufficient muscle strength. The wetness of the oral mucosa is the dryness of the tongue. If it is dry, the movement used for speaking is hindered. In addition, the food is crushed after being taken into the mouth, but it is difficult to swallow directly, so in order to make it easy to swallow the crushed food, saliva plays a role in gathering the crushed food. However, if the mouth is dry, it is difficult to form food blocks (blocks of crushed food gathered). Bite force is the force of biting hard objects, which is the strength of the muscle strength of the lower jaw. Tongue pressure is an indicator of the force of the tongue pushing the palate. If the tongue pressure weakens, there are cases where the movement of swallowing is difficult to carry out. In addition, if the tongue pressure weakens, there are cases where the speed of tongue movement slows down, and there are cases where the speech speed slows down. Chewing function is a comprehensive function of the oral cavity.

[0047] In the present invention, the reduced state of the oral function of the person being evaluated (for example, the reduced state of the elements of the oral function) can be evaluated based on the sound uttered by the person being evaluated. This is because specific features can be seen in the sound uttered by the person being evaluated whose oral function is reduced, and by extracting it as a rhythmic feature, the oral function of the person being evaluated can be evaluated. The present invention is implemented by an oral function evaluation method, a program that enables a computer or the like to execute the method, an oral function evaluation device as an example of the computer, and an oral function evaluation system having the oral function evaluation device. Below, an oral function evaluation system is shown to illustrate the oral function evaluation method, etc.

[0048] [Structure of the oral function evaluation system]

[0049] The configuration of the oral function evaluation system 200 according to the embodiment will be described.

[0050] Figure 1 1 is a diagram showing a configuration of an oral function evaluation system 200 according to an embodiment.

[0051] The oral function evaluation system 200 is a system for evaluating the oral function of the evaluated person U by analyzing the voice of the evaluated person U. Figure 1As shown, an oral function evaluation device 100 and a portable terminal 300 (an example of a terminal) are provided.

[0052] The oral function evaluation device 100 is a device that acquires voice data representing voice uttered by the evaluation subject U through the portable terminal 300 and evaluates the oral function of the evaluation subject U based on the acquired voice data.

[0053] The portable terminal 300 is a sound collection device that collects the sound of the syllable or fixed sentence spoken by the evaluated person U in a non-contact manner, and outputs the sound data representing the collected sound to the oral function evaluation device 100. The syllable or fixed sentence includes a change in the first formant frequency or a change in the second formant frequency and is composed of two or more beats (mora), or includes at least one of a flash, a plosive, a voiceless sound, a glottal sound, and a fricative sound. For example, the portable terminal 300 is a smart phone or a tablet computer with a microphone. In addition, the portable terminal 300 is not limited to a smart phone or a tablet computer as long as it is a device with a sound collection function, and it can also be a notebook PC, etc. In addition, the oral function evaluation system 200 can also be equipped with a sound collection device (microphone) instead of the portable terminal 300. In addition, the oral function evaluation system 200 can also be equipped with an input interface for obtaining personal information of the evaluated person U. The input interface is not particularly limited as long as it has an input function such as a keyboard or a touch. In addition, the volume of the microphone can also be set in the oral function evaluation system 200.

[0054] The portable terminal 300 may also be a display device having a display and displaying an image based on the image data output from the oral function evaluation device 100. That is, the portable terminal 300 is an example of a prompting device for prompting the information output from the oral function evaluation device 100 as an image. In addition, the display device may not be the portable terminal 300, but may be a monitor device composed of a liquid crystal panel or an organic EL panel. That is, in this embodiment, the portable terminal 300 is both a sound collection device and a display device, but the sound collection device (microphone), the input interface and the display device may also be separately provided.

[0055] The oral function evaluation device 100 and the portable terminal 300 may be connected by wire or wirelessly as long as they can transmit and receive audio data or image data for displaying an image showing an evaluation result described later.

[0056] The oral function evaluation device 100 analyzes the voice of the person to be evaluated U based on the voice data collected by the portable terminal 300, evaluates the oral function of the person to be evaluated U based on the analysis result, and outputs the evaluation result. For example, the oral function evaluation device 100 outputs image data for displaying an image representing the evaluation result or data for providing oral advice to the person to be evaluated U generated based on the evaluation result to the portable terminal 300. In this way, the oral function evaluation device 100 can notify the person to be evaluated U of the degree of oral function and advice for preventing the reduction of oral function, so that, for example, the person to be evaluated U can prevent and improve the reduction of oral function.

[0057] The oral function evaluation device 100 is, for example, a personal computer, but may also be a server device. In addition, the oral function evaluation device 100 may also be a portable terminal 300. That is, the portable terminal 300 may also have the functions of the oral function evaluation device 100 described below.

[0058] Figure 2 The oral function evaluation system 200 according to the embodiment is a block diagram showing a characteristic functional structure. The oral function evaluation device 100 includes a voice feature amount calculation device 400 , a calculation unit 130 , an evaluation unit 140 , an output unit 150 , a suggestion unit 160 , and a storage unit 170 .

[0059] The voice feature quantity calculation device 400 is a device that performs calculation by extracting the feature quantity (prosody feature quantity) of the voice of the evaluated person U. Specifically, the voice feature quantity calculation device 400 includes an acquisition unit 110, an S / N ratio calculation unit 115, a sound pressure adjustment unit 116, an extraction unit 120, and an information output unit 180. In addition, an example in which the voice feature quantity calculation device 400 is built into the oral function evaluation device 100 is shown here, but the voice feature quantity calculation device 400 may be provided separately from the oral function evaluation device 100. In this case, the oral function evaluation device 100 may also include an acquisition unit that acquires voice data and personal information, etc., independently of the acquisition unit 110 of the voice feature quantity calculation device 400.

[0060] The acquisition unit 110 acquires voice data obtained by collecting voices spoken by the evaluated person U in a non-contact manner through the portable terminal 300. The voice is a voice in which the evaluated person U speaks a syllable or a fixed sentence pattern that includes a change in the first formant frequency or a change in the second formant frequency and is composed of two or more beats. Alternatively, the voice may be a syllable or a fixed sentence pattern in which at least one of a flash, a plosive, a voiceless, a glottal, and a fricative is spoken. However, in some situations described later, the voice may be a voice in which an arbitrary sentence is spoken. In addition, the acquisition unit 110 may also acquire personal information of the evaluated person U. For example, personal information is information input into the portable terminal 300, such as age, weight, height, sex, BMI (Body Mass Index), dental information (for example, the number of teeth, the presence or absence of dentures, the location of occlusal support, the number of functional teeth, the number of remaining teeth, etc.), serum albumin value, or eating rate, etc. In addition, personal information may be acquired through a swallowing screening tool called EAT-10 (10-item swallowing dysphagia screening form), a Saint-style swallowing questionnaire, a medical interview, the Barthel Index, or a basic checklist, etc. The acquisition unit 110 is, for example, a communication interface for wired or wireless communication.

[0061] The S / N ratio calculation unit 115 is a processing unit that calculates the S / N ratio of the acquired sound data. The S / N ratio of the sound data is the ratio of the second average intensity of the sound collected during the period when the evaluated person U made a sound in the acquired sound data to the first average intensity of the sound collected during the period when the evaluated person U did not make a sound (a period of only dark noise). Therefore, the S / N ratio calculation unit 115 is configured to be able to extract the sound during the period when the evaluated person U did not make a sound from the sound data and calculate the first average intensity, and to extract the sound during the period when the evaluated person U made a sound from the sound data and calculate the second average intensity. Specifically, the S / N ratio calculation unit 115 is implemented by a processor, a microcomputer, or a dedicated circuit.

[0062] The sound pressure adjustment unit 116 is a processing unit that generates and outputs adjusted sound data suitable for evaluating the oral function by adjusting the sound pressure of the sound data when the S / N ratio of the acquired sound data indicates a condition that is not suitable for evaluating the oral function. The adjustment of the sound pressure of the sound data by the sound pressure adjustment unit 116 will be described later. Specifically, the sound pressure adjustment unit 116 is implemented by a processor, a microcomputer, or a dedicated circuit.

[0063] The extraction unit 120 is a processing unit that analyzes the voice data of the evaluated person U acquired by the acquisition unit 110 or the voice data after the sound pressure is adjusted by the sound pressure adjustment unit 116. Specifically, the extraction unit 120 is implemented by a processor, a microcomputer, or a dedicated circuit.

[0064] The extraction unit 120 calculates by extracting the prosodic feature from the sound data acquired by the acquisition unit 110 or output by the sound pressure adjustment unit 116. The prosodic feature is a numerical value representing the feature of the sound of the person to be evaluated U extracted from the sound data and used by the evaluation unit 140 to evaluate the oral function of the person to be evaluated U. The prosodic feature includes a feature related to sound pressure consisting of at least one of a relative difference in sound pressure and a time change in the relative difference in sound pressure. In addition, the prosodic feature may also include at least one of a speech rate, a first formant frequency, a second formant frequency, a change in the first formant frequency, a change in the second formant frequency, a time change in the first formant frequency, a time change in the second formant frequency, an opening time, a closing time, and a time of a plosive.

[0065] The information output unit 180 is a processing unit that outputs information for increasing the S / N ratio. When the calculated S / N ratio does not satisfy a certain standard, the information output unit 180 generates and outputs information such as an instruction to improve the environment for collecting the voice spoken by the evaluated person U. Specifically, the information output unit 180 is implemented by a processor, a microcomputer, or a dedicated circuit.

[0066] The calculation unit 130 calculates the estimated value of the oral function of the evaluation subject U based on the prosodic feature extracted by the extraction unit 120 and a preset estimation formula. Specifically, the calculation unit 130 is implemented by a processor, a microcomputer, or a dedicated circuit.

[0067] The evaluation unit 140 evaluates the deterioration state of the oral function of the evaluated person U by determining the estimated value calculated by the calculation unit 130 using the oral function evaluation index. Index data 172 indicating the oral function evaluation index is stored in the storage unit 170. Specifically, the evaluation unit 140 is implemented by a processor, a microcomputer, or a dedicated circuit.

[0068] The output unit 150 outputs the estimated value calculated by the calculation unit 130 to the suggestion unit 160. In addition, the output unit 150 may also output the evaluation result of the oral function of the evaluated person U evaluated by the evaluation unit 140 to the portable terminal 300 or the like. Specifically, the output unit 150 is implemented by a processor, a microcomputer or a dedicated circuit, and a communication interface for wired communication or wireless communication.

[0069] The suggestion unit 160 compares the estimated value calculated by the calculation unit 130 with the preset data, and makes suggestions related to the oral function of the person to be evaluated U. The suggestion data 173, which is the preset data, is stored in the storage unit 170. In addition, the suggestion unit 160 may also compare the personal information acquired by the acquisition unit 110 with the suggestion data 173, and make suggestions related to the oral cavity to the person to be evaluated U. The suggestion unit 160 outputs the suggestions to the portable terminal 300. The suggestion unit 160 is realized by, for example, a processor, a microcomputer or a dedicated circuit, and a communication interface for wired communication or wireless communication.

[0070] The storage unit 170 is a storage device that stores the estimated formula data 171 representing the estimated formula of the oral function calculated based on a plurality of learning data, the index data 172 representing the oral function evaluation index for determining the estimated value of the oral function of the evaluated person U, the advice data 173 representing the relationship between the estimated value of the oral function and the content of the advice, and the personal information data 174 representing the above-mentioned personal information of the evaluated person U. The estimated formula data 171 is referred to by the calculation unit 130 when calculating the estimated value of the oral function of the evaluated person U. The index data 172 is referred to by the evaluation unit 140 when evaluating the deterioration state of the oral function of the evaluated person U. The advice data 173 is referred to by the advice unit 160 when giving advice related to the oral function to the evaluated person U. The personal information data 174 is, for example, data acquired via the acquisition unit 110. In addition, the personal information data 174 may be stored in the storage unit 170 in advance. The storage unit 170 is realized by, for example, ROM (Read Only Memory), RAM (Random Access Memory), semiconductor memory, HDD (Hard Disk Drive), etc.

[0071] In addition, the storage unit 170 may also store a program executed by a computer to realize each functional unit of the voice feature value calculation device 400, the calculation unit 130, the evaluation unit 140, the output unit 150, and the suggestion unit 160, image data representing the evaluation result used when outputting the evaluation result of the oral function of the evaluated person U, and data such as an image, a moving image, a sound, or a text representing the content of the suggestion. In addition, the storage unit 170 may also store an image for instruction described later.

[0072] Although not shown in the figure, the oral function evaluation device 100 may also include an indication unit for indicating that the person being evaluated U pronounces a syllable or a fixed sentence pattern including a change in the first formant frequency or a change in the second formant frequency and consisting of two or more beats, or including at least one of a flash, a plosive, a voiceless sound, a glottal, and a fricative. Specifically, the indication unit obtains image data of an indication image or audio data of an indication audio stored in the storage unit 170 for indicating the pronunciation of the above syllable or fixed sentence pattern, and outputs the image data or the audio data to the portable terminal 300.

[0073] [Oral function evaluation method processing order]

[0074] Next, a specific processing procedure of the oral function evaluation method executed by the oral function evaluation device 100 will be described.

[0075] Figure 3A This is a flowchart showing a processing procedure for evaluating the oral function of an evaluation subject U based on the oral function evaluation method according to the embodiment. Figure 4 1 is a diagram showing an overview of a method for acquiring the voice of the person to be evaluated U based on the oral function evaluation method.

[0076] First, the instruction unit instructs to pronounce a syllable or a fixed sentence pattern, wherein the syllable or fixed sentence pattern includes a change in the first formant frequency or a change in the second formant frequency and is composed of two or more beats, or includes at least one of a flash, a plosive, a voiceless sound, a glottal, and a fricative sound (step S101). For example, in step S101, the instruction unit obtains image data of an image for indicating to the evaluated person U stored in the storage unit 170, and outputs the image data to the portable terminal 300. Then, as Figure 4 As shown in (a) of FIG. 1 , an image for indicating to the evaluated person U is displayed on the portable terminal 300. Figure 4In (a), an example of a fixed phrase is "I decided to draw a picture". However, you can also use the indicative phrases "The old man who made the flowers bloom and the monkey and the crab compete", "Draw a picture of the flowers", "The sunflowers have bloomed", etc. In addition, you can also use the indicative phrases "Ippai", "Ittai", "Ikkai", "Pattan", "Kappa", "Tail", "Kikkari", "Only", etc. In addition, you can also use the indicative phrases "kara", "sara", "chara", "jara", "shara", "kyara", "pura", etc. In addition, you can also give instructions to say "aei", "iea", "ai", "ia", "kakeki", "kikeka", "naneni", "chiteta", "papepi", "pipepapa", "katepi", "chipeka", "kaki", "tachi", "papi", "misa", "rari", "wani", "niwa", "eo", "io", "iu", "teko", "kiro", "teru", "peko", "memo", "emo", etc. Pronunciation instructions can also be repeated to give instructions to say such pronunciations.

[0077] In addition, the instruction section can also be passed through to obtain the existing information section 170 in the middle of the user's direction. However, it is also possible not to use the vocal image and voice for instructions, and it is the idea that the recipient (family member, medical student, etc.) can direct the user's own voice to the user's own oral function.

[0078] For example, the spoken syllables or fixed sentence patterns may include two or more vowels, or a combination of vowels and consonants that are accompanied by the opening and closing of the mouth or the forward and backward movement of the tongue for speaking. For example, as such a syllable or fixed sentence pattern, Japanese has "えをかくことにきめたよ" and so on. In order to say "えを" in "えをかくことにきめたよ", the tongue moves back and forth, and in order to say "きめた" in "えをかくことにきめたよ", the mouth opens and closes. The "えを" part of "えをかくことにきめたよ" includes the second formant frequencies of the vowel "e" and the vowel "o", and because the vowel "e" and the vowel "o" are adjacent, it includes the change amount of the second formant frequency. In addition, the time change of the second formant frequency is included in this part. The "きめた" part of "えをかくことにきめたよ" includes the first formant frequencies of the vowel "i", the vowel "e", and the vowel "a", and because the vowel "i", the vowel "e", and the vowel "a" are adjacent, it includes the change amount of the first formant frequency. In addition, the time change of the first formant frequency is included in this part. By saying "えをかくことにきめたよ", prosodic features such as relative sound pressure difference, first formant frequency, second formant frequency, change in first formant frequency, change in second formant frequency, time change in first formant frequency, time change in second formant frequency, and speech rate can be extracted.

[0079] For example, the fixed sentence pattern may include the repetition of syllables composed of a flash sound and a consonant different from the flash sound. For example, Japanese has such a fixed sentence pattern as "からからから..." By repeatedly saying "からからから...", it is possible to extract prosodic features such as relative difference in sound pressure, temporal change in relative difference in sound pressure, temporal change in sound pressure, and number of repetitions.

[0080] For example, the spoken syllable or fixed sentence pattern may include at least one combination of a vowel and a plosive. For example, Japanese has such a syllable, such as "いったい". By speaking "いったい", prosodic features such as the relative difference in sound pressure and the time of the plosive (the time between vowels) can be extracted.

[0081] In addition, since the rhythmic feature of the relative sound pressure difference is easily affected by dark noise, the rhythmic feature of the relative sound pressure difference may have an adverse effect on the accuracy of the estimation of the estimated value, especially in the case of a sound collection environment with a relatively small S / N ratio. Therefore, in the present invention, the sound pressure of the sound data is adjusted according to the S / N ratio calculated in the S / N ratio calculation unit 115 so that the calculated (extracted) feature of the relative sound pressure difference becomes appropriate. In the present invention, in this way, the appropriate rhythmic feature of the relative sound pressure difference can be calculated, and the estimated value can be estimated by reducing the possibility that the inappropriate rhythmic feature of the relative sound pressure difference will have an adverse effect on the accuracy in the estimation of the estimated value.

[0082] use Figure 3B to Figure 3H The specific processing and other operations for this will be described. Figure 3B This is a flowchart showing a processing procedure for sound data used in the oral function evaluation method according to the embodiment. Figure 3C This is a diagram showing an example of information output in the oral function evaluation method according to the embodiment. Figure 3D This is a flowchart showing a procedure for processing voice data used in an oral function evaluation method according to another example of the embodiment. Figure 3E to Figure 3G It is a diagram for explaining adjustment of the sound pressure of audio data according to the embodiment. Figure 3H This is a graph showing the relationship between the processing of adjusting the sound pressure and the accuracy (estimation accuracy) in the oral function evaluation method according to the embodiment.

[0083] like Figure 3B As shown, the S / N ratio calculation unit 115 measures the dark noise for calculating the S / N ratio, and calculates the first average intensity (sound pressure) of only the dark noise (step S201). In the measurement of the dark noise, the sound during the period when the person being evaluated U does not make a sound can be extracted for use. For example, as described above, when the person being evaluated U speaks the instructed syllable or fixed sentence, the sound can be extracted during the period of only dark noise before and after the syllable or fixed sentence. If there is a part where the sound is interrupted in the fixed sentence, the sound can be extracted as a period of only dark noise at that part.

[0084] Next, the S / N ratio calculation unit 115 calculates the second average intensity (sound pressure) of the person being evaluated U when speaking in order to calculate the S / N ratio (step S202). At this time, the sound when the instructed syllable or fixed sentence is spoken can be used, or an instruction can be given to make the person speak any syllable or fixed sentence separately in order to collect the sound. Alternatively, if the person being evaluated U is talking to someone before the oral function evaluation is about to be performed, the first average intensity and the second average intensity can also be calculated using this situation.

[0085] Furthermore, the S / N ratio calculation unit 115 calculates the S / N ratio by calculating the ratio of the second average intensity to the first average intensity (step S203). Here, the calculated S / N ratio is output to the information output unit 180. Next, the information output unit 180 determines whether the S / N ratio is greater than the second threshold value (step S204). When it is determined that the S / N ratio is below the second threshold value ("No" in S204), the information output unit 180 generates and outputs information on improving the sound collection environment to increase the S / N ratio (step S205).

[0086] For example, in Figure 3C As an example of the case where such information is output, the state where "Please check the connection status of the microphone or increase the volume when speaking" is displayed on the portable terminal 300 is shown. By outputting information in this way, an instruction to increase the S / N ratio by at least one of reducing the dark noise, that is, reducing the first average intensity, and increasing the volume when speaking, that is, increasing the second average intensity, is given. In addition, the portable terminal 300 may be caused to display "Please move the sound collection location" to reduce the ambient sound when the person being evaluated speaks.

[0087] In addition, as another example with the same effect, you can also execute Figure 3D The process shown in Figure 3D In, relative to Figure 3B , which is different in that step S204a is executed after step S201 and before step S202 instead of step S204, but the other points are the same. Here, the S / N ratio is not calculated, and the presence or absence of information output is determined only based on the size of the dark noise. Specifically, in step S204a, it is determined whether the first average intensity is smaller than the sound pressure threshold. When the first average intensity is less than the sound pressure threshold ("Yes" in S204), step S202 and step S203 are executed, and the process proceeds to step S206. On the other hand, when the first average intensity is greater than the sound pressure threshold ("No" in S204), the process proceeds to step S205, and information on improving the sound collection environment for the information output unit 180 to increase the S / N ratio is generated and output. In this example, it is different from the example in that the presence or absence of information output can be determined only based on the first average intensity without calculating the S / N ratio. Figure 3B That is, the advantage is that the quality of the sound collection environment can be determined from the viewpoint of the noise amount before instructing the evaluation subject U to speak in order to calculate the second average intensity.

[0088] Back to Figure 3B , when it is determined that the S / N ratio is greater than the second threshold value ("Yes" in S204) (or Figure 3DAfter step S203), the information output unit 180 does nothing in particular and proceeds to step S206. Specifically, the calculated S / N ratio is also output to the sound pressure adjustment unit 116. The sound pressure adjustment unit 116 determines whether the S / N ratio is greater than the first threshold value (step S206). In the case where it is determined that the S / N ratio is less than the first threshold value ("No" in S206), the sound pressure adjustment unit 116 adjusts the sound pressure of the sound data (step S208), and the processing ends. On the other hand, in the case where it is determined that the S / N ratio is greater than the first threshold value ("Yes" in S206), the sound pressure adjustment unit 116 does not adjust the sound pressure of the sound data (step S207), and the processing ends.

[0089] In this way, the sound pressure of the audio data is adjusted (or not adjusted) according to the S / N ratio, and the audio data is provided for extraction of the prosodic feature value.

[0090] Here, use Figure 3E to Figure 3F This section explains how to adjust the sound pressure. Figure 3E , the intensity change of the sound data with respect to time when the S / N ratio is less than the second threshold value (or the first average intensity is greater than the sound pressure threshold value) is shown. Figure 3E In the example of , since the value of the S / N ratio is so small that even if the sound pressure is adjusted, the appropriate prosody feature cannot be extracted, so the information is output from the information output unit 180 to improve the sound collection environment. Figure 3E In the case of the voice data shown, neither the prosodic feature value extraction nor the oral cavity function evaluation is performed.

[0091] exist Figure 3F In the figure, the intensity change of the sound data relative to time (upper section) and the change of the fundamental frequency (pitch) relative to time (lower section) are shown when the S / N ratio is greater than the second threshold value and less than the first threshold value. Figure 3F In the example of , the sound pressure is adjusted, and the sound data after the sound pressure is adjusted (dashed line) is generated for the acquired sound data (solid line). As shown in the figure, the sound pressure adjustment here is performed at the timing when the sound pressure of the sound data is at a minimum value and the fundamental frequency is 0 (the part of the hollow arrow in the figure). In this way, the sound pressure at the timing when there is no sound and the intensity of the sound is extremely small can be appropriately adjusted.

[0092] Furthermore, in the adjustment of the sound pressure, the sound pressure of the difference between the intensity of the sound in silence and the first average intensity is subtracted from the sound pressure of the above-mentioned timed minimum value. As a result, no major changes occur around the maximum part in the sound data, only the minimum value decreases, and the influence of dark noise is reduced. In this way, the relative difference in sound pressure characterized by the difference between the maximum value and the minimum value is extracted as a more appropriate feature quantity. In addition, regarding the intensity of the sound in silence mentioned above, the pre-set hypothetical intensity can be stored in the storage unit 170, etc., and read out for use when adjusting the sound pressure, or the lowest value of the intensity actually measured in the past under the same sound collection conditions can be used as the intensity of the sound in silence.

[0093] On the other hand, Figure 3G , the intensity change of the audio data with respect to time when the S / N ratio is greater than or equal to the first threshold value is shown. Figure 3G In the example, the solid line represents an example of sound data of a healthy person, and the dotted line represents an example of sound data of a symptomatic person. When the S / N ratio is greater than the first threshold, if the sound pressure is adjusted randomly, the sound data of the symptomatic person may be misjudged as the sound data of a healthy person. Therefore, it is effective to prohibit the adjustment of the sound pressure when the S / N ratio is greater than the first threshold.

[0094] Because of the above, the second threshold value can be determined by experience or experiment to be set to a value larger than the S / N ratio at which the appropriate prosodic feature cannot be extracted even if the sound pressure is adjusted. In addition, the first threshold value can be determined by experience or experiment to not adjust the sound pressure of the voice data of the symptomatic person. For example, Figure 3H In the figure, (a) shows the relationship between the S / N ratio and the estimation accuracy when the prosodic feature is extracted from the original sound data without considering the S / N ratio, and (b) shows the relationship between the S / N ratio and the estimation accuracy when the prosodic feature is extracted from the sound data whose sound pressure is adjusted according to the S / N ratio.

[0095] like Figure 3H As shown in FIG. 1 , if the S / N ratio is greater than the first threshold, the estimation accuracy is the same in (a) and (b). However, in the range where the S / N ratio is less than the first threshold and greater than the second threshold, Figure 3H In (a), the rhythmic feature quantity related to sound pressure affected by dark noise reduces the estimation accuracy, which is different from Figure 3H The inference accuracy will be lower than that of (b). Figure 3HAs shown by the single-dot chain line in (b), if the S / N ratio is less than the second threshold, an instruction to increase the S / N ratio is given. Therefore, before the estimated value is estimated, the environment is transferred to an improved S / N ratio and the processing is performed again from the acquisition of the sound data. Therefore, it is not easy to estimate the estimated value in a state of low estimation accuracy. However, in a state where the S / N ratio is less than the second threshold, there is also an estimation accuracy ratio. Figure 3H In the case where (a) is high, it may be useful to estimate the estimated value even in this state.

[0096] Back to Figure 3A As explained in the example, the voice data can also be obtained by collecting the voice of the evaluated person U speaking a syllable or a fixed sentence at least twice at different speech speeds. For example, the evaluated person U is instructed to speak "えをかくことにきめたよ" at a normal speed and a speed faster than that. By speaking "えをかくことにきめたよ" at a normal speed and a speed faster than that, the degree of maintenance of the state of the oral function can be estimated.

[0097] Then, if Figure 3A As shown in FIG. 1 , the acquisition unit 110 acquires the voice data of the evaluated person U who has received the instruction in step S101 via the portable terminal 300 (step S102). Figure 4 As shown in (b) of FIG. 1 , in step S102 , for example, the evaluated person U utters a syllable or a fixed sentence pattern such as “えをかくことにきめたよ” toward the portable terminal 300. The acquisition unit 110 acquires the syllable or fixed sentence pattern uttered by the evaluated person U as voice data.

[0098] Next, the extraction unit 120 extracts the prosody feature from the sound data acquired by the acquisition unit 110 when the S / N ratio is greater than the first threshold value, and extracts the prosody feature from the sound data output by the sound pressure adjustment unit 116 when the S / N ratio is less than the first threshold value (step S103).

[0099] For example, when the sound data acquired by the acquisition unit 110 is obtained from the sound of saying "えをかくことにきめたよ", the extraction unit 120 extracts the relative difference in sound pressure, the first formant frequency, the second formant frequency, the amount of change in the first formant frequency, the amount of change in the second formant frequency, the time change of the first formant frequency, the time change of the second formant frequency, and the speech rate as prosodic feature quantities. Figure 5A and Figure 5B This is explained.

[0100] Figure 5A This is a diagram showing an example of voice data showing the voice of the evaluee U speaking "えをかくことにきめたよ". Figure 5A The horizontal axis of the graph shown is time, and the vertical axis is power (sound pressure). Figure 5A The unit of power shown on the vertical axis of the graph is decibel (dB).

[0101] exist Figure 5A In the graph shown, the changes in sound pressure corresponding to "え", "を", "か", "く", "こ", "と", "に", "き", "め", "た", and "よ" can be confirmed. Obtain Department 110 at Figure 3A In step S102 shown in the figure, the user U is obtained Figure 5A The S / N ratio calculation unit calculates the S / N ratio of the acquired sound data, and the sound data provided to the extraction unit 120 is determined to be either the sound data acquired by the acquisition unit 110 or the sound data output by the sound pressure adjustment unit 116 based on the calculated S / N ratio.

[0102] The extraction unit 120, for example, Figure 3A In step S103 shown in the figure, the extraction is performed by a known method. Figure 5A The sound data shown includes the sound pressures of "k" and "a" in "か(ka)", the sound pressures of "k" and "o" in "こ(ko)", the sound pressures of "t" and "o" in "と(to)", and the sound pressures of "t" and "a" in "た(ta)". The extraction unit 120 extracts the relative difference Diff_P(ka) of the sound pressures of "k" and "a" as a prosodic feature based on the extracted sound pressures of "k" and "a". Similarly, the extraction unit 120 extracts the relative difference Diff_P(ko) of the sound pressures of "k" and "o", the relative difference Diff_P(to) of the sound pressures of "t" and "o", and the relative difference Diff_P(ta) of the sound pressures of "t" and "a" as prosodic features. For example, based on the relative difference in sound pressure, the oral function related to the force of swallowing (the pressure of the tongue contacting the palate) or the force of gathering food can be evaluated. Furthermore, based on the relative difference in sound pressure including "k", the oral function related to the ability to prevent food and drink from flowing into the throat can be evaluated.

[0103] Figure 5B A diagram showing an example of change in the formant frequency of the voice of the evaluee U saying "えをかくことにきめたよ". Specifically, Figure 5B This is a graph for explaining an example of changes in the first formant frequency and the second formant frequency.

[0104] The first resonance peak frequency is the peak frequency of the amplitude that can be seen first from the low frequency side of the human voice, and is known to easily reflect the characteristics related to the opening and closing of the mouth. The second resonance peak frequency is the peak frequency of the amplitude that can be seen second from the low frequency side of the human voice, and is known to easily reflect the influence related to the forward and backward movement of the tongue.

[0105] The extraction unit 120 extracts the first formant frequency and the second formant frequency of each of the plurality of vowels as prosodic feature quantities from the voice data representing the voice spoken by the evaluator U. For example, the extraction unit 120 extracts the second formant frequency F2e corresponding to the vowel "e" in "えを" and the second formant frequency F2o corresponding to the vowel "o" as prosodic feature quantities. In addition, for example, the extraction unit 120 extracts the first formant frequency F1i corresponding to the vowel "i" in "きめた", the first formant frequency F1e corresponding to the vowel "e", and the first formant frequency F1a corresponding to the vowel "a" as prosodic feature quantities.

[0106] Furthermore, the extraction unit 120 extracts the variation of the first formant frequency and the variation of the second formant frequency of the continuous vowel character string as the prosodic feature quantity. For example, the extraction unit 120 extracts the variation of the second formant frequency F2e and the second formant frequency F2o (F2e-F2o), and the variation of the first formant frequency F1i, the first formant frequency F1e, and the first formant frequency F1a (F1e-F1i, F1a-F1e, F1a-F1i) as the prosodic feature quantity.

[0107] Furthermore, the extraction unit 120 extracts the time variation of the first formant frequency and the time variation of the second formant frequency of the continuous vowel string as the prosodic feature. For example, the extraction unit 120 extracts the time variation of the second formant frequency F2e and the second formant frequency F2o, and the time variation of the first formant frequency F1i, the first formant frequency F1e, and the first formant frequency F1a as the prosodic feature. Figure 5B , an example of the time change of the first formant frequency F1i, the first formant frequency F1e, and the first formant frequency F1a is shown in FIG. 1 , and the time change is ΔF1 / ΔTime. ΔF1 is F1a-F1i.

[0108] For example, based on the second resonance peak frequency, the amount of change in the second resonance peak frequency, or the time change in the second resonance peak frequency, the oral function related to the movement of gathering food (the front, back, left, and right movement of the tongue) can be evaluated. In addition, for example, based on the first resonance peak frequency, the amount of change in the first resonance peak frequency, or the time change in the first resonance peak frequency, the oral function related to the ability to crush food can be evaluated. In addition, based on the time change in the first resonance peak frequency, the oral function related to the ability to move the mouth faster can be evaluated.

[0109] In addition, if Figure 5AAs shown, the extraction unit 120 may also extract the speech rate as the prosodic feature. For example, the extraction unit 120 may also extract the time from when the evaluated person U starts to say "えをかくことにきめたよ" to when the person finishes saying it as the prosodic feature. In addition, for example, the extraction unit 120 is not limited to the time from when the entire "えをかくことにきめたよ" is said to be finished, and the extraction unit 120 may also extract the time from when a specific part of "えをかくことにきめたよ" is said to be finished as the prosodic feature. In addition, for example, the extraction unit 120 may also extract the average time taken to say one word or a plurality of words of the entire or specific part of "えをかくことにきめたよ" as the prosodic feature. For example, based on the speech rate, oral functions related to swallowing movements, movements to bring food together, or tongue dexterity can be evaluated.

[0110] For example, when the speech data acquired by the acquisition unit 110 is speech data obtained by repeatedly saying "からからから...", the extraction unit 120 extracts the temporal change of the relative difference in sound pressure as the prosodic feature value. Figure 6 This is explained.

[0111] Figure 6 1 is a diagram showing an example of audio data showing the audio of the evaluated person U repeatedly saying “Kara Kara Kara…”. Figure 6 The horizontal axis of the graph shown is time, and the vertical axis is power (sound pressure). Figure 6 The unit of power shown on the vertical axis of the graph is decibel (dB).

[0112] exist Figure 6 In the graph shown in FIG. 1 , the change in sound pressure corresponding to “か” and “ら” can be confirmed. Figure 3A In step S102 shown in the figure, the user U is obtained Figure 6 The extraction unit 120, for example, Figure 3A In step S103 shown in the figure, the extraction is performed by a known method. Figure 6The sound data shown includes the sound pressures of "k" and "a" in "か(ka)", and the sound pressures of "r" and "a" in "ら(ra)". The extraction unit 120 extracts the relative difference Diff_P(ka) between the sound pressures of "k" and "a" as a prosodic feature based on the extracted sound pressures of "k" and "a". Similarly, the extraction unit 120 extracts the relative difference Diff_P(ra) between the sound pressures of "r" and "a" as a prosodic feature. For example, the extraction unit 120 extracts the relative difference Diff_P(ka) and the relative difference Diff_P(ra) as prosodic features for each of the repeatedly spoken "から". Next, the extraction unit 120 extracts the temporal variation of the relative sound pressure difference Diff_P(ka) as a prosodic feature quantity based on the extracted relative sound pressure differences Diff_P(ka), and extracts the temporal variation of the relative sound pressure difference Diff_P(ra) as a prosodic feature quantity based on the extracted relative sound pressure differences Diff_P(ra). For example, based on the temporal variation of the relative sound pressure difference, the oral function related to the movement of swallowing, the movement of gathering food, or the ability to crush food can be evaluated.

[0113] In addition, the extraction unit 120 may extract the temporal change of the sound pressure as the prosodic feature. For example, the temporal change of the minimum sound pressure (the sound pressure of "k") in each "kara" when "kara kara..." is repeatedly said may be extracted, the temporal change of the maximum sound pressure (the sound pressure of "a") in each "kara" may be extracted, and the temporal change of the sound pressure (the sound pressure of "r") between "kara" and "ra" in each "kara" may be extracted. For example, based on the temporal change of the sound pressure, the oral function related to the swallowing movement, the movement of gathering food, or the ability to crush food can be evaluated.

[0114] In addition, if Figure 6 As shown, the extraction unit 120 may also extract the number of times "から" is uttered per a predetermined time, that is, the number of repetitions as a feature quantity. The predetermined time is not particularly limited, and is 5 seconds, etc. For example, based on the number of repetitions per a predetermined time, the oral function related to the movement of swallowing or the movement of gathering food can be evaluated.

[0115] For example, when the sound data acquired by the acquisition unit 110 is sound data obtained from the sound of saying "いったい", the extraction unit 120 extracts the relative difference in sound pressure and the time of the plosive sound as the prosodic feature amount. Figure 7 Explain it.

[0116] Figure 7 : is a figure which shows an example of the audio data which shows the sound which the evaluation subject U said "hittai". Here, an example of the audio data which shows the sound which repeatedly said "hittaihittai..." is shown. Figure 7The horizontal axis of the shown curve graph is time, and the vertical axis is power (sound pressure). Additionally, Figure 7 the unit of the power shown on the vertical axis of the curve graph is decibel (dB).

[0117] In Figure 7 the shown curve graph, it is possible to confirm the changes in sound pressure corresponding to "i", "tsu", "ta", "i". Acquisition unit 110, in Figure 3A the shown step S102, acquires from the evaluatee U Figure 7 the shown sound data. Extraction unit 120, for example, in Figure 3A the shown step S103, extracts, by a known method, Figure 7 the sound pressures of "t" and "a" in "ta" included in the shown sound data. Extraction unit 120 extracts the relative sound pressure difference Diff_P(ta) of "t" and "a" as a prosodic feature quantity according to the extracted sound pressures of "t" and "a". For example, according to the relative sound pressure difference, it is possible to evaluate the oral functions related to the swallowing force or the force for gathering food. In addition, extraction unit 120 extracts the time Time(i - ta) of the plosive sound (the time of the plosive sound between "i" and "ta") as a prosodic feature quantity. For example, according to the time of the plosive sound, it is possible to evaluate the oral functions related to the swallowing movement, the movement for gathering food, or the stable movement of the tongue.

[0118] In addition, syllables or fixed phrases in Japanese are listed as examples of spoken syllables or fixed phrases, but it is not limited to Japanese, and any language is acceptable.

[0119] Figure 8 is a diagram showing an example of Chinese syllables or fixed phrases that are similar to Japanese syllables or fixed phrases in terms of the movement of the tongue or the degree of opening and closing of the mouth during pronunciation.

[0120] There are various languages in the world, but there are languages with similar degrees of movement of the tongue or opening and closing of the mouth during pronunciation. For example, when pronouncing the Chinese "高岛屋打卡即刻打一五折" (hereinafter denoted as gao dao wu da ka ji ke dayi wu zhe), the degree of movement of the tongue or opening and closing of the mouth during pronunciation is similar to that when pronouncing the Japanese "えをかくことにきめたよ", so it is possible to extract prosodic feature quantities similar to those of the Japanese "えをかくことにきめたよ". Additionally, the notation of tone marks is omitted in this specification. In Figure 8 several examples of syllables or fixed phrases with similar degrees of movement of the tongue or opening and closing of the mouth during pronunciation in Japanese and Chinese are shown for reference only.

[0121] In addition, use Fig. 9A and Fig. 9B , a brief explanation is given of the various languages ​​in the world that have similar tongue movements or the degree of opening and closing of the mouth during pronunciation.

[0122] Fig.9A It is a diagram showing the International Phonetic Alphabet of vowels.

[0123] Fig. 9B It is a diagram showing the International Phonetic Alphabet of consonants.

[0124] Fig. 9A In the positional relationship of the International Phonetic Symbols of the vowels shown, the horizontal direction represents the forward and backward movement of the tongue. The closer the tongue is, the more similar the forward and backward movement of the tongue is. The vertical direction represents the degree of opening and closing of the mouth. The closer the tongue is, the more similar the degree of opening and closing of the mouth is. Fig. 9B In the table of international phonetic symbols of consonants shown, the horizontal direction indicates the parts used for pronunciation from lips to throat, and the international phonetic symbols in the same grid of the table can be used to pronounce the same sound using the same part. Therefore, the present invention can be applied to various languages ​​existing in the world.

[0125] For example, if you want to increase the opening and closing of the mouth, Fig. 9A The international phonetic symbols (e.g., "i" and "a") shown as being separated in the vertical direction are included in a syllable or a fixed sentence. As a result, the variation of the first formant frequency can be increased as a prosodic feature. In addition, for example, when the front and rear positions of the tongue are to be increased, Fig. 9A The international phonetic symbols (such as "i" and "u") that are separated in the horizontal direction are included in a syllable or a fixed sentence pattern. This makes it possible to increase the variation of the second formant frequency as a prosodic feature.

[0126] For example, when the sound data acquired by the acquisition unit 110 is sound data obtained from the sound of saying "gao dao wu da ka ji ke da yiwu zhe", the extraction unit 120 extracts the relative difference in sound pressure, the first formant frequency, the second formant frequency, the change in the first formant frequency, the change in the second formant frequency, the time change of the first formant frequency, the time change of the second formant frequency, and the speech rate as prosodic feature quantities. Fig. 10A and Fig. 10B This is explained.

[0127] Fig. 10A 2 is a diagram showing an example of audio data showing the audio of the evaluated person U saying “gao dao wu da ka ji ke da yi wu zhe”. Fig. 10A The horizontal axis of the graph shown is time, and the vertical axis is power (sound pressure). Fig. 10AThe unit of power shown on the vertical axis of the graph is decibel (dB).

[0128] exist Fig. 10A In the graph shown in FIG. 1 , the changes in sound pressure corresponding to “gao”, “dao”, “wu”, “da”, “ka”, “ji”, “ke”, “da”, “yi”, “wu”, and “zhe” can be confirmed. Figure 3A In step S102 shown in the figure, the user U is obtained Fig. 10A The extraction unit 120, for example, Figure 3A In step S103 shown in the figure, the extraction is performed by a known method. Fig. 10A The sound data shown includes the sound pressures of "d" and "a" in "dao", the sound pressures of "k" and "a" in "ka", the sound pressures of "k" and "e" in "ke", and the sound pressures of "zh" and "e" in "zhe". The extraction unit 120 extracts the relative difference Diff_P(da) between the sound pressures of "d" and "a" as a prosodic feature based on the extracted sound pressures of "d" and "a". Similarly, the extraction unit 120 extracts the relative difference Diff_P(ka) between the sound pressures of "k" and "a", the relative difference Diff_P(ke) between the sound pressures of "k" and "e", and the relative difference Diff_P(zhe) between the sound pressures of "zh" and "e" as prosodic feature. For example, based on the relative difference in sound pressure, the oral function related to the force of swallowing or the force of gathering food can be evaluated. Furthermore, based on the relative difference in sound pressure including "k", the oral function related to the ability to prevent food and drink from flowing into the throat can be evaluated.

[0129] Fig. 10B : is a diagram showing an example of changes in the formant frequency of the sound of the evaluated person U saying "gao dao wu da ka ji ke da yi wu zhe". Specifically, Fig. 10B This is a graph for explaining an example of changes in the first formant frequency and the second formant frequency.

[0130] The extraction unit 120 extracts the first formant frequency and the second formant frequency of each of the plurality of vowels as prosodic feature quantities from the voice data representing the voice spoken by the evaluator U. For example, the extraction unit 120 extracts the first formant frequency F1i corresponding to the vowel "i" in "ji", the first formant frequency F1e corresponding to the vowel "e" in "ke", and the first formant frequency F1a corresponding to the vowel "a" in "da" as prosodic feature quantities. In addition, for example, the extraction unit 120 extracts the second formant frequency F2i corresponding to the vowel "i" in "yi" and the second formant frequency F2u corresponding to the vowel "u" ​​in "wu" as prosodic feature quantities.

[0131] Furthermore, the extraction unit 120 extracts the variation of the first formant frequency and the variation of the second formant frequency of the character string of consecutive vowels as prosodic feature quantities. For example, the extraction unit 120 extracts the variation of the first formant frequency F1i, the first formant frequency F1e, and the first formant frequency F1a (F1e-F1i, F1a-F1e, F1a-F1i), and the variation of the second formant frequency F2i and the second formant frequency F2u (F2i-F2u) as prosodic feature quantities.

[0132] Furthermore, the extraction unit 120 extracts the time variation of the first formant frequency and the time variation of the second formant frequency of the continuous vowel character string as the prosodic feature quantity. For example, the extraction unit 120 extracts the time variation of the first formant frequency F1i, the first formant frequency F1e, and the first formant frequency F1a, and the time variation of the second formant frequency F2i and the second formant frequency F2u as the prosodic feature quantity.

[0133] For example, based on the second resonance peak frequency, the amount of change in the second resonance peak frequency, or the time change in the second resonance peak frequency, the oral function related to the movement of gathering food can be evaluated. In addition, based on the first resonance peak frequency, the amount of change in the first resonance peak frequency, or the time change in the first resonance peak frequency, the oral function related to the ability to crush food can be evaluated. In addition, based on the time change in the first resonance peak frequency, the oral function related to the ability to move the mouth faster can be evaluated.

[0134] In addition, if Fig. 10A As shown, the extraction unit 120 may also extract the speech rate as the prosodic feature. For example, the extraction unit 120 may also extract the time from when the person being evaluated U starts to say "gao dao wu da ka ji ke da yi wu zhe" to when the person being evaluated finishes saying it as the prosodic feature. In addition, for example, it is not limited to the time from when the entire "gao dao wu da ka ji ke da yi wu zhe" is said to be finished. The extraction unit 120 may also extract the time from when a specific part of "gao dao wu da ka ji ke dayi wu zhe" is said to be finished as the prosodic feature. In addition, for example, the extraction unit 120 may also extract the average time taken to say one word or a plurality of words of the entire or specific part of "gao dao wu da ka ji ke da yi wu zhe" as the prosodic feature. For example, based on the speech rate, it is possible to evaluate the oral function related to the movement of swallowing, the movement of gathering food, or the dexterity of the tongue.

[0135] Back to Figure 3AAs described in , the calculation unit 130 calculates the estimated value of the oral function of the evaluated person U based on the extracted prosodic feature value and the estimation formula of the oral function calculated based on a plurality of learning data (step S104).

[0136] The inference formula of oral function is set in advance based on the evaluation results of multiple subjects. The sound feature quantities spoken by the subjects are collected, and the oral function of the subjects is actually diagnosed. The correlation between the sound feature quantities and the diagnosis results is set through statistical analysis such as multiple regression. Different types of inference formulas can be generated according to the selection method of the sound feature quantities used as representative values. In this way, the inference formula can be generated in advance.

[0137] Furthermore, in order to express the correlation between the sound feature quantity and the diagnosis result, machine learning may be used for setting. Machine learning methods include logistic regression, SVM (Support Vector Machine), random forest, and the like.

[0138] For example, the estimation formula may be configured to include coefficients corresponding to elements of oral function and variables into which the extracted prosodic feature values ​​are substituted and multiplied by the coefficients. The following equations 1 to 5 are examples of estimation formulas.

[0139] Estimated value of tongue coating adhesion = (A1×F2e)+(B1×F2o)+(C1×F1i)+(D1×F1e)+(E1×F1a)+(F1×Diff_P(ka))+(G1×Diff_P(ko))+(H1×Diff_P(to))+(J1×Diff_P(ta))+(K1×Diff_P(ka))+(L1×Diff_P(ra))+(M1×Num(kara))+(N1×Diff_P(ta))+(P1×Time(i-ta))+Q1 (Formula 1)

[0140] Estimated value of oral mucosal moisture = (A2×F2e)+(B2×F2o)+(C2×F1i)+(D2×F1e)+(E2×F1a)+(F2×Diff_P(ka))+(G2×Diff_P(ko))+(H2×Diff_P(to))+(J2×Diff_P(ta))+(K2×Diff_P(ka))+(L2×Diff_P(ra))+(M2×Num(kara))+(N2×Diff_P(ta))+(P2×Time(i-ta))+Q2 (Formula 2)

[0141] Estimated value of bite force = (A3×F2e)+(B3×F2o)+(C3×F1i)+(D3×F1e)+(E3×F1a)+(F3×Diff_P(ka))+(G3×Diff_P(ko))+(H3×Diff_P(to))+(J3×Diff_P(ta))+(K3×Diff_P(ka))+(L3×Diff_P(ra))+(M3×Num(kara))+(N3×Diff_P(ta))+(P3×Time(i-ta))+Q3 (Formula 3)

[0142] Estimated value of tongue pressure = (A4×F2e)+(B4×F2o)+(C4×F1i)+(D4×F1e)+(E4×F1a)+(F4×Diff_P(ka))+(G4×Diff_P(ko))+(H4×Diff_P(to))+(J4×Diff_P(ta))+(K4×Diff_P(ka))+(L4×Diff_P(ra))+(M4×Num(kara))+(N4×Diff_P(ta))+(P4×Time(i-ta))+Q4 (Formula 4)

[0143] Estimated value of chewing function = (A5×F2e)+(B5×F2o)+(C5×F1i)+(D5×F1e)+(E5×F1a)+(F5×Diff_P(ka))+(G5×Diff_P(ko))+(H5×Diff_P(to))+(J5×Diff_P(ta))+(K5×Diff_P(ka))+(L5×Diff_P(ra))+(M5×Num(kara))+(N5×Diff_P(ta))+(P5×Time(i-ta))+Q5 (Formula 5)

[0144] A1, B1, C1, ..., P1, A2, B2, C2, ..., P2, A3, B3, C3, ..., P3, A4, B4, C4, ..., P4, A5, B5, C5, ..., P5 are coefficients, specifically, coefficients corresponding to elements of oral function. For example, A1, B1, C1, ..., P1 are coefficients corresponding to tongue coating adhesion, which is one of the elements of oral function, A2, B2, C2, ..., P2 are coefficients corresponding to oral membrane wetness, which is one of the elements of oral function, A3, B3, C3, ..., P3 are coefficients corresponding to bite force, which is one of the elements of oral function, A4, B4, C4, ..., P4 are coefficients corresponding to tongue pressure, which is one of the elements of oral function, and A5, B5, C5, ..., P5 are coefficients corresponding to chewing function, which is one of the elements of oral function.

[0145] Q1 is a constant corresponding to the degree of tongue coating adhesion, Q2 is a constant corresponding to the wetness of the oral membrane, Q3 is a constant corresponding to the bite force, Q4 is a constant corresponding to the tongue pressure, and Q5 is a constant corresponding to the chewing function.

[0146] F2e multiplied by A1, A2, A3, A4, A5 and F2o multiplied by B1, B2, B3, B4, B5 are variables into which the second formant frequency as the prosodic feature quantity extracted from the utterance data when the evaluated person U utters “えをかくことにきめたよ” is substituted. F1i multiplied by C1, C2, C3, C4, C5, F1e multiplied by D1, D2, D3, D4, D5 and F1a multiplied by E1, E2, E3, E4, E5 are variables into which the first formant frequency as the prosodic feature quantity extracted from the utterance data when the evaluated person U utters “えをかくことにきめたよ” is substituted. Diff_P(ka) multiplied by F1, F2, F3, F4, F5, Diff_P(ko) multiplied by G1, G2, G3, G4, G5, Diff_P(to) multiplied by H1, H2, H3, H4, H5, and Diff_P(ta) multiplied by J1, J2, J3, J4, J5 are variables substituted as the relative difference in sound pressure of the prosodic feature extracted from the speech data when the person being evaluated U utters "えをかくことにきめたよ". Diff_P(ka) multiplied by K1, K2, K3, K4, K5 and Diff_P(ra) multiplied by L1, L2, L3, L4, L5 are variables into which the relative difference in sound pressure of the prosodic feature extracted from the speech data when the evaluated person U utters "から". Num(kara) multiplied by M1, M2, M3, M4, M5 is a variable into which the number of repetitions of the prosodic feature extracted from the speech data when the evaluated person U repeatedly utters "から" within a certain period is substituted. Diff_P(ta) multiplied by N1, N2, N3, N4, N5 is a variable into which the relative difference in sound pressure of the prosodic feature extracted from the speech data when the evaluated person U utters "いったい". Time (i-ta) multiplied by P1, P2, P3, P4, and P5 is a variable into which the time of the plosive sound, which is a prosodic feature extracted from the utterance data when the evaluated person U utters “いったい”, is substituted.

[0147] As shown in the above formulas 1 to 5, for example, the calculation unit 130 calculates the estimated value for each element of the oral function of the evaluated person U (for example, tongue coating adhesion, oral mucosal moisturizing degree, bite force, tongue pressure, and chewing function). In addition, these elements of oral function are examples, and the elements of oral function only need to include at least one of the tongue coating adhesion, oral mucosal moisturizing degree, bite force, tongue pressure, cheek pressure, number of remaining teeth, swallowing function, and chewing function of the evaluated person U.

[0148] In addition, for example, the extraction unit 120 extracts a plurality of prosodic features from voice data obtained by collecting voices of the evaluated person U speaking a plurality of types of syllables or fixed sentence patterns (for example, "えをかくことにきめたよ", "から", and "いったい" in the above-mentioned Formulas 1 to 5), and the calculation unit 130 calculates the estimated value of the oral function based on the extracted plurality of prosodic features and the estimation formula. The calculation unit 130 can calculate the estimated value of the oral function with high accuracy by substituting the plurality of prosodic features extracted from the voice data of the plurality of types of syllables or fixed sentence patterns into one estimation formula.

[0149] Furthermore, although a linear equation is shown as the estimated equation, the estimated equation may be a multi-level equation such as a quadratic equation.

[0150] Next, the evaluation unit 140 evaluates the reduced state of the oral function of the evaluated person U by using the oral function evaluation index to determine the estimated value calculated by the calculation unit 130 (step S105). For example, the evaluation unit 140 evaluates the reduced state of the oral function of the evaluated person U according to each element of the oral function by using the oral function evaluation index set for each element of the oral function to determine the estimated value of each element of the oral function. The oral function evaluation index is an index used to evaluate the oral function, for example, a condition for determining that the oral function is reduced. Fig.11 The oral function evaluation index is described.

[0151] Fig.11 This is a diagram showing an example of oral function evaluation index.

[0152] Oral function evaluation indexes are set for each element of oral function. For example, an index of 50% or more is set for tongue coating adhesion, an index of 27 or less is set for oral mucosal moistness, an index of less than 500N is set for bite force (when using Dental Prescale II of GC Co., Ltd.), an index of less than 30kPa is set for tongue pressure, and an index of less than 100mg / dL is set for chewing function (for indexes, refer to "Basic Considerations on Oral Function Decrease" (https: / / www.jads.jp / basic / pdf / document_02.pdf) of the Japanese Dental Association). The evaluation unit 140 evaluates the state of oral function decline of the person being evaluated U according to each element of oral function by comparing the calculated estimated value of each element of oral function with the oral function evaluation index set for each element of oral function. For example, when the calculated estimated value of the tongue coating adhesion is greater than 50%, the tongue coating adhesion, which is an element of the oral function, is evaluated to be in a reduced state. Similarly, when the calculated estimated value of the oral mucosal moisturizing is less than 27, the oral mucosal moisturizing, which is an element of the oral function, is evaluated to be in a reduced state. When the calculated estimated value of the bite force is less than 500N, the bite force, which is an element of the oral function, is evaluated to be in a reduced state. When the calculated estimated value of the tongue pressure is less than 30kPa, the tongue pressure, which is an element of the oral function, is evaluated to be in a reduced state. When the calculated estimated value of the chewing function is less than 100mg / dL, the chewing function, which is an element of the oral function, is evaluated to be in a reduced state. In addition, as oral function evaluation indicators set for the tongue coating adhesion, oral mucosal moisturizing, bite force, tongue pressure and chewing function, Fig.11 What is shown is an example, and is not limited to this. For example, the index of remaining teeth can also be set for chewing function. In addition, tongue coating adhesion, oral mucosal moistness, bite force, tongue pressure and chewing function are shown as elements of oral function, but these are examples. For example, for reduced tongue, mouth and lip movement function, there are elements of oral function such as tongue movement, lip movement, lip strength, etc.

[0153] Back to Figure 3A As described in , the output unit 150 outputs the evaluation result of the oral function of the evaluated person U evaluated by the evaluation unit 140 (step S106). For example, the output unit 150 outputs the evaluation result to the portable terminal 300. In this case, the output unit 150 may also include a communication interface for wired communication or wireless communication, obtain image data of an image corresponding to the evaluation result from the storage unit 170, and send the obtained image data to the portable terminal 300. Fig.12 and Fig.13 An example of the image data (evaluation result) is shown in FIG.

[0154] Fig.12 and Fig.13 : is a diagram showing an example of the evaluation results of each element of oral function. Fig.12 As shown, the evaluation result may be an evaluation result of two levels, OK or NG. OK means normal, and NG means abnormal. In addition, normal and abnormal may not be indicated for each element of the oral function. For example, the evaluation result may only indicate the element with reduced suspicion. In addition, the evaluation result is not limited to the evaluation result of two levels, and may be a more detailed evaluation result in which the degree of evaluation is divided into three or more levels. In this case, the indicator data 172 stored in the storage unit 170 may include multiple indicators for one element. In addition, Fig.13 The evaluation results are shown in a radar chart. Fig.12 and Fig.13 In the data, oral cleanliness, food holding power, force of biting hard objects, tongue strength, and movement of the jaw are shown as elements of oral function. Regarding oral cleanliness, the evaluation result is presented based on the degree of tongue coating adhesion; regarding food holding power, the evaluation result is presented based on the moistness of the oral mucosa; regarding the force of biting hard objects, the evaluation result is presented based on the bite force; regarding tongue strength, the evaluation result is presented based on tongue pressure; and regarding the movement of the jaw, the evaluation result is presented based on the estimated value of the chewing function. In addition, Fig.12 and Fig.13 This is an example, and the evaluation item sentences, oral function items, and their corresponding combinations are not limited to Fig.12 and Fig.13 The content shown.

[0155] Back to Figure 3A As described in , the suggestion unit 160 compares the estimated value calculated by the calculation unit 130 with the pre-set data (suggestion data 173) to make suggestions related to the oral function of the evaluated person U (step S107). Fig.14 Describe the preset data.

[0156] Fig.14 This is an example of preset data (advice data 173 ) used when providing advice on oral function.

[0157] like Fig.14As shown, the recommendation data 173 is data that establishes a correspondence between the evaluation result and the recommendation content for each element of the oral function. For example, when the estimated value of the calculated oral cleanliness is less than 50%, the recommendation unit 160 determines that it is OK because the index is met, and makes a recommendation based on the recommendation content established in correspondence with the oral cleanliness. In addition, the specific recommendation content is omitted, but for example, the storage unit 170 includes data (for example, images, moving images, sounds, texts, etc.) indicating the recommendation content, and the recommendation unit 160 uses such data to make recommendations related to the oral function to the evaluated person U.

[0158] [Effects, etc.]

[0159] As described above, the sound feature quantity calculation method related to the first technical solution of the present invention is a sound feature quantity calculation method executed by a computer, and is used to calculate the rhythmic feature quantity (feature quantity) of the sound of the evaluated person U based on the sound spoken by the evaluated person U, obtain sound data obtained by collecting the sound spoken by the evaluated person U; adjust the sound pressure of the sound data based on the first average intensity of the sound collected during the period when the evaluated person U did not make any sound in the acquired sound data; and calculate the rhythmic feature quantity including at least the feature quantity related to the sound pressure based on the sound data after the sound pressure is adjusted.

[0160] Thus, when the feature quantity related to sound pressure calculated based on the voice spoken by the evaluated person U is not suitable for use as it is, for example, in the evaluation of oral function, etc., the feature quantity related to sound pressure can be calculated as an appropriate prosodic feature quantity based on the voice data after the sound pressure is adjusted. Thus, the feature quantity related to sound pressure calculated is more appropriate from the viewpoint of, for example, the evaluation of the oral function of the evaluated person U. That is, the prosodic feature quantity of the voice can be calculated more appropriately based on the voice of the evaluated person.

[0161] In addition, for example, the sound feature quantity calculation method related to the second technical solution may further calculate an S / N ratio as a ratio of a second average intensity to a first average intensity in the sound feature quantity calculation method described in the first technical solution, and when the calculated S / N ratio is below a first threshold value, the sound pressure of the sound data is adjusted, the second average intensity being the average intensity of the sound collected during the period when the person being evaluated utters the sound in the acquired sound data.

[0162] Thus, it is possible to determine whether the feature quantity related to sound pressure is not suitable for use as it is, based on the condition of whether the S / N ratio is below the first threshold. According to this determination, the sound pressure of the voice data can be adjusted, and the feature quantity related to sound pressure can be calculated based on the voice data after the sound pressure is adjusted. Thus, the feature quantity related to sound pressure calculated is more appropriate from the perspective of, for example, evaluating the oral function of the person being evaluated U. That is, the prosodic feature quantity of the voice can be more appropriately calculated based on the voice of the person being evaluated.

[0163] Furthermore, for example, the sound feature value calculation method according to the third aspect may be such that, in the sound feature value calculation method according to the second aspect, when the calculated S / N ratio is larger than the first threshold value, the sound pressure of the sound data is not adjusted.

[0164] Thus, it is possible to determine whether the feature quantity related to sound pressure is suitable for use as it is, based on the condition of whether the S / N ratio is greater than the first threshold value. According to this determination, it is possible to calculate the feature quantity related to sound pressure based on the sound data with the sound pressure as it is, without adjusting the sound pressure of the sound data. Thus, the feature quantity related to sound pressure calculated is more appropriate from the perspective of, for example, evaluating the oral function of the person being evaluated U. That is, the prosodic feature quantity of the voice can be more appropriately calculated based on the voice of the person being evaluated.

[0165] In addition, for example, the sound feature quantity calculation method related to the fourth technical solution may be, in the sound feature quantity calculation method recorded in any one of the first to third technical solutions, adjust the sound pressure of the sound data by subtracting the sound pressure of the difference between the intensity of the sound in silence measured in advance and the first average intensity from the sound pressure of the sound data.

[0166] Therefore, by calculating the difference in sound pressure between the intensity of the sound in silence that has been measured in advance and that corresponds to the original sound pressure in silence and the intensity of the sound when the person being evaluated U does not speak under the condition that the sound in the sound data is collected, and subtracting the difference, the sound pressure can be adjusted so that the intensity of the sound when the person being evaluated U does not speak becomes the intensity corresponding to the original sound pressure in silence.

[0167] Furthermore, for example, the sound feature amount calculation method according to the fifth technical solution may be such that, in the sound feature amount calculation method described in any one of the first to fourth technical solutions, the sound pressure of the sound data is adjusted at the timing when the sound pressure of the sound data reaches a minimum value.

[0168] Thus, it is possible to determine that the person being evaluated U is not uttering a sound based on the timing at which the sound pressure reaches a minimum value, and to adjust the sound pressure.

[0169] Furthermore, for example, the sound feature amount calculation method according to the sixth aspect may be such that, in the sound feature amount calculation method according to the fifth aspect, the sound pressure of the sound data is adjusted at a timing when the fundamental frequency in the sound data becomes 0.

[0170] Thus, it is possible to determine that the person being evaluated U is not uttering a sound based on the timing when the sound pressure becomes 0, and to adjust the sound pressure.

[0171] In addition, for example, the sound feature quantity calculation method related to the 7th technical scheme may be a sound feature quantity calculation method recorded in any one of the 1st to 6th technical schemes, wherein the feature quantity includes: at least one of the relative sound pressure difference and the change in the relative sound pressure difference; and at least one of the resonance peak, the resonance peak change, the opening time, the closing time, the time of the plosive sound and the speaking rate.

[0172] Thus, in the calculated feature quantities, at least one of the relative sound pressure difference and the change in the relative sound pressure difference can be included as feature quantities related to sound pressure, and at least one of the resonance peak, resonance peak change, opening time, closing time, time of plosive sound and speech rate can be included as other feature quantities.

[0173] Furthermore, for example, the sound feature value calculation method according to the eighth technical solution may further output information for increasing the S / N ratio when the calculated S / N ratio is below the second threshold value in the sound feature value calculation method described in the second or third technical solution.

[0174] Thus, when it is desired to calculate the prosody feature value based on speech data collected in an unsuitable environment where the S / N ratio is less than the second threshold value which is smaller than the first threshold value, it is possible to take measures to improve the environment. Calculation of the prosody feature value based on speech data collected in such an unsuitable environment is suppressed.

[0175] In addition, for example, the sound feature calculation method related to the 9th technical solution may also output information for increasing the S / N ratio when the first average intensity is above the sound pressure threshold in the sound feature calculation method recorded in any one of the 2nd or 3rd technical solutions.

[0176] Thus, when it is desired to calculate the prosody feature value from the speech data collected in an unsuitable environment where the first average intensity is equal to or greater than the sound pressure threshold, measures can be taken to improve the environment. Calculation of the prosody feature value from the speech data collected in such an unsuitable environment is suppressed.

[0177] In addition, for example Figure 3AAs shown, it may also include: an acquisition step (step S102), acquiring sound data obtained by collecting the sound of the evaluated person U speaking a syllable or a fixed sentence pattern, wherein the syllable or fixed sentence pattern includes a change in the first resonance peak frequency or a change in the second resonance peak frequency and is composed of more than 2 beats, or includes at least one of a flash, a plosive, a voiceless sound, a glottal and a fricative sound; an extraction step (step S103), extracting a rhythmic feature from the acquired sound data; a calculation step (step S104), calculating an estimated value of the oral function of the evaluated person U based on an estimated formula of the oral function and the extracted rhythmic feature, wherein the estimated formula of the oral function is calculated based on multiple learning data; and an evaluation step (step S105), evaluating the reduced state of the oral function of the evaluated person U by using an oral function evaluation index to determine the calculated estimated value.

[0178] Thus, by obtaining sound data suitable for the evaluation of oral function, the oral function of the person U to be evaluated can be easily evaluated. That is, the oral function of the person U to be evaluated can be evaluated only by the person U to be evaluated speaking the above-mentioned syllables or fixed sentences toward a sound collecting device such as the portable terminal 300. In particular, since the estimated value of the oral function is calculated using an inference formula calculated based on a plurality of learning data, the reduced state of the oral function can be quantitatively evaluated. In addition, since the oral function is not evaluated by directly comparing the rhythmic feature value with the threshold value, but the inference value is calculated based on the rhythmic feature value and the inference formula, and the inference value is compared with the threshold value (oral function evaluation index), the reduced state of the oral function can be evaluated with high accuracy.

[0179] For example, the estimation formula may include coefficients corresponding to elements of oral function and variables into which the extracted prosodic feature values ​​are substituted and multiplied by the coefficients.

[0180] Thus, the estimated value of the oral cavity function can be easily calculated simply by substituting the extracted prosodic feature into the estimation formula.

[0181] For example, in the calculation step, an estimated value is calculated for each element of the oral function of the evaluated person U; in the evaluation step, the calculated estimated value of each element of the oral function is determined using an oral function evaluation index set for each element of the oral function, thereby evaluating the reduced state of the oral function of the evaluated person U for each element of the oral function.

[0182] Thus, the state of oral function decline can be evaluated for each element. For example, by preparing an estimation formula for each element of oral function, in which the coefficient differs according to the element of oral function, the state of oral function decline can be easily evaluated for each element.

[0183] For example, the oral function elements may include at least one of the tongue coating adhesion, oral mucosal moistness, bite force, tongue pressure, cheek pressure, number of remaining teeth, swallowing function, and chewing function of the evaluated person U.

[0184] Thus, the decline state of at least one oral function element of the evaluated person U, including tongue coating adhesion, oral mucosal moistness, bite force, tongue pressure, cheek pressure, number of remaining teeth, swallowing function and chewing function, can be evaluated.

[0185] For example, the prosodic feature may include at least one of speech rate, relative sound pressure difference, time change of relative sound pressure difference, first formant frequency, second formant frequency, change of first formant frequency, change of second formant frequency, time change of first formant frequency, time change of second formant frequency and time of plosive sound.

[0186] Since pronunciation changes due to decreased oral function, the decreased state of oral function can be evaluated based on these prosodic feature quantities.

[0187] For example, in the extraction step, multiple prosodic features may be extracted from voice data obtained by collecting voices of the evaluated person U speaking multiple types of syllables or fixed sentences; and in the calculation step, an inference value may be calculated based on the extracted multiple prosodic features and an inference formula.

[0188] Thus, by using a plurality of prosodic feature quantities extracted based on a plurality of types of syllables or fixed sentence patterns for one estimation formula, it is possible to improve the calculation accuracy of the estimation value of the oral cavity function.

[0189] For example, a syllable or a fixed sentence pattern may include two or more vowels or a combination of vowels and consonants, which are accompanied by the opening and closing of the mouth or the forward and backward movement of the tongue in order to speak.

[0190] Thus, prosodic features including the change in the first formant frequency, the time change in the first formant frequency, the change in the second formant frequency, or the time change in the second formant frequency can be extracted from the sound of the evaluated person U speaking such syllables or fixed sentences.

[0191] For example, the voice data may be obtained by collecting the voices of the evaluated person U speaking a syllable or a fixed sentence pattern at least twice at different speech speeds.

[0192] Therefore, the degree to which the state of the oral function is maintained can be estimated based on the sound of the evaluated person U uttering such syllables or fixed sentence patterns.

[0193] For example, the fixed sentence pattern may include repetition of a syllable consisting of a flap and a consonant different from the flap.

[0194] Thus, prosodic feature quantities including temporal changes in the relative difference in sound pressure, temporal changes in sound pressure, and the number of repetitions can be extracted from the sound of the evaluated person U speaking such syllables or fixed sentence patterns.

[0195] For example, a syllable or a fixed sentence pattern may include at least one combination of a vowel and a plosive sound.

[0196] Thus, it is possible to extract prosodic feature quantities including the relative difference in sound pressure and the timing of plosive sounds from the sound of the evaluated person U speaking such syllables or fixed sentence patterns.

[0197] For example, the oral function evaluation method may further include an advice step of providing advice on the oral function of the person to be evaluated U by comparing the calculated estimated value with preset data.

[0198] Thus, the person being evaluated U can receive advice on what kind of countermeasures should be taken when the oral function is reduced.

[0199] The sound feature quantity calculation device 400 related to the 10th technical solution of the present disclosure is a sound feature quantity calculation device 400 that calculates the feature quantity of the sound of the evaluated person U based on the sound spoken by the evaluated person U, and comprises: an acquisition unit 110, which acquires sound data obtained by collecting the sound spoken by the evaluated person U; a sound pressure adjustment unit 116, which adjusts the sound pressure of the sound data based on the first average intensity of the sound collected during the period when the evaluated person U did not make any sound in the acquired sound data; and an extraction unit 120, which calculates the feature quantity by extracting at least the feature quantity related to the sound pressure from the sound data after the sound pressure is adjusted.

[0200] This can produce the same effects as the above-described method for calculating the sound feature amount.

[0201] The oral function evaluation device 100 related to the 11th technical solution of the present disclosure comprises: the sound feature quantity calculation device 400 recorded in the 10th technical solution; a calculation unit 130, which calculates the estimated value of the oral function of the evaluated person U based on the inference formula of the feature quantity related to the sound pressure among the feature quantities extracted from the sound data, and the feature quantity extracted from the sound data after the sound pressure is adjusted; and an evaluation unit 140, which uses the oral function evaluation index to determine the calculated inference value, thereby evaluating the reduced state of the oral function of the evaluated person U.

[0202] In this way, the oral function of the person to be evaluated U can be evaluated using the voice feature amount calculation device 400 .

[0203] For example, the device may further include: a sound collecting device (microphone) used for collecting the voice spoken by the person to be evaluated U; and a presentation device (portable terminal 300) for presenting the deterioration state of the oral function of the person to be evaluated U.

[0204] In addition, for example, it may also include: an acquisition unit 110, which acquires sound data obtained by collecting sounds of the evaluated person U speaking syllables or fixed sentences, wherein the syllables or fixed sentences include changes in the first resonance peak frequency or changes in the second resonance peak frequency and are composed of more than two beats, or include at least one of a flash sound, a plosive sound, a voiceless sound, a glottal sound, and a fricative sound; an extraction unit 120, which extracts prosodic features from the acquired sound data; a calculation unit 130, which calculates an estimated value of the oral function of the evaluated person U based on an estimated formula of the oral function and the extracted prosodic features, wherein the estimated formula of the oral function is calculated based on multiple learning data; and an evaluation unit 140, which uses an oral function evaluation index to determine the calculated estimated value, thereby evaluating the reduced state of the oral function of the evaluated person U.

[0205] Thus, the oral function evaluation device 100 capable of easily evaluating the oral function of the person to be evaluated U can be provided.

[0206] Furthermore, for example, the device may include the oral function evaluation device 100 and a sound collecting device (portable terminal 300 ) to collect the sound of the person to be evaluated U speaking syllables or fixed sentence patterns in a non-contact manner.

[0207] Thus, the oral function evaluation system 200 capable of easily evaluating the oral function of the person to be evaluated U can be provided.

[0208] (Other embodiments)

[0209] As mentioned above, although the oral function evaluation method etc. concerning embodiment were demonstrated, this invention is not limited to the said embodiment.

[0210] For example, the candidate of the inference formula may be updated based on the evaluation result obtained when the expert actually diagnoses the oral function of the person to be evaluated U. In this way, the evaluation accuracy of the oral function can be improved. Machine learning may also be used to improve the evaluation accuracy of the oral function.

[0211] In addition, for example, regarding the suggestion data 173, the suggestion content may be evaluated by the evaluated person U, and updated based on the evaluation result. For example, in the case where a suggestion is made regarding an oral function that is not a problem for the evaluated person U, the evaluated person U evaluates the suggestion content as wrong. And, by updating the suggestion data 173 based on the evaluation result, the wrong suggestion as described above is not made. In this way, the suggestion content related to the oral function for the evaluated person U can be made more effective. In addition, in order to make the suggestion content related to the oral function more effective, machine learning can also be used.

[0212] In addition, for example, the evaluation results of oral function may be stored together with personal information as big data and used for machine learning. In addition, the content of advice related to oral function may be stored together with personal information as big data and used for machine learning.

[0213] In addition, for example, in the above-mentioned embodiment, the oral function evaluation method includes the advice step (step S107 ) of giving advice on the oral function, but this step may not be included. In other words, the oral function evaluation device 100 may not include the advice unit 160 .

[0214] In addition, for example, in the above-described embodiment, the acquisition step (step S102 ) acquires the personal information of the person to be evaluated U, but this information may not be acquired. In other words, the acquisition unit 110 may not acquire the personal information of the person to be evaluated U.

[0215] In addition, for example, the steps in the oral function evaluation method can also be executed by a computer (computer system). Furthermore, the present invention can also be implemented as a program for causing a computer to execute the steps included in these methods. Furthermore, the present invention can also be implemented as a computer-readable non-temporary recording medium such as a CD-ROM recording the program.

[0216] For example, when the present invention is implemented by a program (software), each step is performed by executing the program using hardware resources such as a CPU, a memory, and an input / output circuit of a computer. That is, each step is performed by the CPU obtaining data from a memory or an input / output circuit and performing calculations, or outputting calculation results to a memory or an input / output circuit.

[0217] Furthermore, each component included in the oral function evaluation device 100 and the oral function evaluation system 200 of the above-described embodiment may be realized as a dedicated or general-purpose circuit.

[0218] Furthermore, each component included in the oral function evaluation device 100 and the oral function evaluation system 200 of the above-described embodiment may be realized as an LSI (Large Scale Integration) which is an integrated circuit (IC).

[0219] Furthermore, the integrated circuit is not limited to LSI, and may be realized by a dedicated circuit or a general-purpose processor. A programmable FPGA (Field Programmable Gate Array) or a reconfigurable processor that can reconfigure the connection and setting of circuit cells inside the LSI may also be used.

[0220] Furthermore, if integrated circuit technology that replaces LSI emerges due to advances in semiconductor technology or other derived technologies, it is of course possible to integrate the components included in the oral function evaluation device 100 and the oral function evaluation system 200 using this technology.

[0221] In addition, various modifications that can be conceived by those skilled in the art to the embodiments and embodiments achieved by arbitrarily combining the components and functions of the embodiments without departing from the gist of the present invention are also included in the present invention.

[0222] Description of Reference Numerals

[0223] 100 Oral function evaluation device

[0224] 110 Acquisition Department

[0225] 115S / N ratio calculation unit

[0226] 116 Sound pressure adjustment unit

[0227] 120 Extraction Department

[0228] 130 Computing Department

[0229] 140 Evaluation Department

[0230] 150 Output

[0231] 160 Recommendation Department

[0232] 180 Information output unit

[0233] 200 Oral Function Evaluation System

[0234] 300 portable terminal (terminal, microphone, prompt device)

[0235] 400 Sound feature quantity calculation device

[0236] U Evaluated

Claims

1. A method for calculating a voice feature quantity, executed by a computer, for calculating a feature quantity of a voice of an evaluated person based on a voice spoken by the evaluated person, wherein: Acquiring voice data obtained by collecting the voices spoken by the above-mentioned person being evaluated, adjusting the sound pressure of the sound data based on a first average intensity of sounds collected during a period when the person being evaluated did not make any sound in the acquired sound data, The feature quantity including at least a feature quantity related to the sound pressure is calculated based on the sound data after the sound pressure is adjusted.

2. The method for calculating the sound feature quantity according to claim 1, wherein: further calculating an S / N ratio as a ratio of a second average intensity to the first average intensity, the second average intensity being an average intensity of sounds collected during a period in which the evaluated person uttered the sound in the acquired sound data, When the calculated S / N ratio is equal to or less than a first threshold value, the sound pressure of the audio data is adjusted.

3. The method for calculating the sound feature quantity according to claim 2, wherein: When the calculated S / N ratio is greater than the first threshold value, the sound pressure of the audio data is not adjusted.

4. The method for calculating the sound feature quantity according to claim 1, wherein: In the adjustment of the sound pressure of the sound data, the sound pressure is adjusted by subtracting the sound pressure of the difference between the intensity of the sound in silence measured in advance and the first average intensity from the sound pressure of the sound data.

5. The method for calculating the sound feature quantity according to claim 1, wherein: The sound pressure of the sound data is adjusted at the timing when the sound pressure in the sound data reaches a minimum value.

6. The method for calculating the sound feature quantity according to claim 5, wherein: The sound pressure of the sound data is adjusted at a timing when the fundamental frequency in the sound data becomes 0.

7. The method for calculating the sound feature quantity according to claim 1, wherein: The feature quantities include: at least one of a relative sound pressure difference and a change in the relative sound pressure difference; and at least one of a formant, a change in the formant, an opening time, a closing time, a time of a plosive sound, and a speech rate.

8. The method for calculating the sound feature quantity according to claim 2 or 3, wherein: When the calculated S / N ratio is equal to or lower than the second threshold value, information for increasing the S / N ratio is further outputted.

9. The method for calculating the sound feature quantity according to claim 2 or 3, wherein: When the first average intensity is equal to or greater than the sound pressure threshold, information for increasing the S / N ratio is further output.

10. A voice feature quantity calculation device, which calculates the feature quantity of the voice of the evaluated person based on the voice spoken by the evaluated person, wherein: have: an acquisition unit that acquires voice data obtained by collecting voices spoken by the person to be evaluated; a sound pressure adjustment unit that adjusts the sound pressure of the sound data based on a first average intensity of sounds collected during a period when the person being evaluated does not make any sound in the acquired sound data; as well as The extraction unit calculates the feature amount by extracting the feature amount including at least the feature amount related to the sound pressure from the sound data after the sound pressure is adjusted.

11. An oral function evaluation device, wherein: have: The sound feature quantity calculation device according to claim 10; a calculation unit that calculates an estimated value of the oral function of the subject based on an estimation formula that includes, in calculation, a feature quantity related to the sound pressure among the feature quantities extracted from the voice data and the feature quantity extracted from the voice data after adjusting the sound pressure; as well as The evaluation unit determines the calculated estimated value using an oral function evaluation index to evaluate the deterioration state of the oral function of the person to be evaluated.

Images