JP2018050847A - Cognitive function evaluation apparatus, cognitive function evaluation method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a cognitive function evaluation device, a cognitive function evaluation method, and a program capable of obtaining highly accurate evaluation results of cognitive function in a short time. SOLUTION: In a cognitive function evaluation device 1, a formant analysis unit 13 receives target data representing a time variation of an instantaneous sound pressure of a specific phoneme included in a subject's voice over a target period. Then, the formant analysis unit 13 divides the target period into a plurality of frames, and obtains the frequency of a specific formant for each of the two or more target frames. The feature analysis unit 14 obtains a feature amount for a specific formant frequency obtained for each target frame. The evaluation unit 15 evaluates the cognitive function of the subject based on the feature amount. [Selection diagram] Fig. 1

Description

  The present invention relates to a cognitive function evaluation apparatus, a cognitive function evaluation method, and a program.

  Conventionally, a technique has been proposed in which prosodic feature values are extracted from speech data, and the risk level of cognitive impairment is calculated for the speaker of the speech data from the prosodic feature values of the speech data (see, for example, Patent Document 1). In the technique of Patent Document 1 (hereinafter referred to as conventional technique), a learning data storage unit stores a plurality of types of prosodic feature quantities extracted from speech data as learning data in advance. Then, the weight determination unit learns and determines the weight for each of the speech prosodic feature values in the learning model using the learning data. The feature quantity extraction unit calculates a risk level of cognitive impairment by extracting a plurality of types of prosodic feature quantities from the input speech data and inputting the extracted plurality of types of prosodic feature quantities into a learning model.

  In the prior art, the prosodic feature value is a feature value related to the frequency component of speech, a feature value related to the formant structure of speech, a feature value related to the volume of speech, a feature value related to speech speed, and a feature value related to the response time until answering a question Etc. are used.

JP 2011-255106 A

  For evaluation of cognitive function, there is a demand for shortening the evaluation time and improving the evaluation accuracy.

  Therefore, an object of the present invention is to provide a cognitive function evaluation apparatus, a cognitive function evaluation method, and a program that can obtain a highly accurate evaluation result of a cognitive function in a short time.

  A cognitive function evaluation apparatus according to an aspect of the present invention includes a formant analysis unit, a feature analysis unit, and an evaluation unit. The formant analysis unit receives target data representing a temporal variation of instantaneous sound pressure of a specific phoneme included in a target person's voice over a target period. The formant analysis unit divides the target period of the target data into a plurality of frames, and sets at least a specific formant frequency generated in the speech for each of two or more target frames of the plurality of frames. Ask. The feature analysis unit obtains a feature amount for the frequency of the specific formant obtained for each target frame. The evaluation unit evaluates the cognitive function of the subject based on the feature amount.

The cognitive function evaluation method according to an aspect of the present invention includes the following steps.
The formant analysis unit receives target data representing the temporal variation of the instantaneous sound pressure of a specific phoneme included in the target person's voice over the target period. Then, the formant analysis unit divides the target period for the target data into a plurality of frames, and sets at least a specific formant frequency generated in the speech for each of two or more target frames of the plurality of frames. Step to seek.
A step in which a feature analysis unit obtains a feature amount for the frequency of the specific formant obtained for each target frame;
The evaluation unit evaluates the cognitive function of the target person based on the feature amount.

  A program according to one embodiment of the present invention causes a computer to function as a cognitive function evaluation device. The program causes the computer to function as a formant analysis unit, a feature analysis unit, and an evaluation unit. The formant analysis unit receives target data representing a temporal variation of instantaneous sound pressure of a specific phoneme included in a target person's voice over a target period. The formant analysis unit divides the target period of the target data into a plurality of frames, and sets at least a specific formant frequency generated in the speech for each of two or more target frames of the plurality of frames. Ask. The feature analysis unit obtains a feature amount for the frequency of the specific formant obtained for each target frame. The evaluation unit evaluates the cognitive function of the subject based on the feature amount.

  In this invention, there exists an effect that the evaluation result of a highly accurate cognitive function can be obtained in a short time.

FIG. 1 is a block diagram illustrating a system configuration including a cognitive function evaluation apparatus according to an embodiment. FIG. 2 is a spectrum diagram of the above-mentioned voice. FIG. 3A is a diagram showing the distribution of formants of voices uttered by a person with reduced cognitive function. FIG. 3B is a diagram showing a distribution of formants of voices uttered by the healthy person. FIG. 4 is a flowchart showing the processing of the cognitive function evaluation apparatus according to the above. FIG. 5 is an explanatory diagram for explaining the processing of the formant analysis unit of the cognitive function evaluation apparatus. FIG. 6A is a diagram showing another distribution of sound formants emitted by a person with reduced cognitive function. FIG. 6B is a diagram showing another distribution of sound formants emitted by the healthy person. FIG. 7 is a block diagram showing the configuration of the cognitive function evaluation apparatus in Modification 1 of the above. FIG. 8 is a block diagram showing the configuration of the cognitive function evaluation apparatus in Modification 2 of the above. FIG. 9 is a block diagram showing a configuration of a robot provided with the above cognitive function evaluation apparatus. FIG. 10 is a block diagram showing a configuration of a nurse call system provided with the above cognitive function evaluation apparatus.

  The present embodiment relates to a cognitive function evaluation apparatus, a cognitive function evaluation method, and a program. More specifically, the present embodiment relates to a cognitive function evaluation device, a cognitive function evaluation method, and a program that evaluate a person's cognitive function based on a person's utterance.

  The cognitive function evaluation apparatus described below evaluates cognitive function based on the formant of a person's voice. Therefore, the cognitive function evaluation apparatus can evaluate a person's cognitive function without understanding the utterance content.

  In the present embodiment, evaluation of human cognitive function is evaluation of the degree (progression degree) of MCI (Mild Cognitive Impairment) and evaluation of the degree (progression degree) of cognitive dysfunction.

  Although the target person of cognitive function evaluation mainly assumes an elderly person, the person who is suspected of juvenile Alzheimer's disease may be sufficient. Elderly people who are subject to cognitive function evaluation include elderly people who live in welfare facilities, elderly people who use day service centers, elderly people who are living alone, or those who live in housing for elderly people with services. Assumed. Hereinafter, the target person of the cognitive function evaluation is simply referred to as the target person.

  FIG. 1 shows a configuration example of the cognitive function evaluation system 10. The cognitive function evaluation system 10 includes a cognitive function evaluation apparatus 1, a microphone 21, and a communication interface 22. Hereinafter, the communication interface 22 is abbreviated as a communication I / F 22.

  The cognitive function evaluation device 1 includes an acquisition unit 11, an extraction unit 12, a formant analysis unit 13, a feature analysis unit 14, an evaluation unit 15, an evaluation notification unit 16, and an object identification unit 17.

  The microphone 21 collects the subject's voice and converts it into an electrical signal, and outputs the electrical signal as analog voice data. That is, the audio data output from the microphone 21 is an analog electrical signal.

  Here, FIG. 2 shows an example of the result of spectrum analysis of human voice data. In FIG. 2, a curve Y1 indicates an envelope of the power spectrum.

  A human voice has a formant whose power peaks at a specific frequency, and the formant is a feature that distinguishes the voice. For example, when spectrum analysis is performed on audio data, the frequencies f1, f2, f3,. . . In each of (f1 <f2 <f3 <...), Formants F1, F2, F3,. . . Is present. Formants F1, F2, F3,. . . Of the first formant F1, the second formant F2, the third formant F3,. . . Called. In the following, the frequencies f1, f2, f3,. . . , Formant frequencies f1, f2, f3,. . . Call it.

  The first formant F1 is considered to depend mainly on the opening of a person's mouth. The second formant F2 is considered to depend mainly on the position and movement of the human tongue. The third formant F3 is considered to depend mainly on the shape and movement of the human throat.

  Here, it is known that when a cognitive function is lowered, a function of a working memory (working memory) in the brain is lowered. Working memory is related to the work of the human anterior lobe. Then, when the function of the working memory is lowered, it tends to be difficult for a person to perform two or more operations at the same time. That is, it becomes difficult to talk while moving or talking while thinking. As a result, when the function of the working memory is lowered, when the movement of speaking is combined with another movement, the movement of the tongue becomes dull or the position of the tongue during speaking cannot be determined. This appears as a symptom that, for example, the utterance operation in which a person speaks stops spinning.

  Since the second formant F2 mainly depends on the position and movement of the person's tongue, it is considered that the decrease in the function of the working memory affects the second formant F2. That is, a new finding has been obtained that when the cognitive function is lowered, the degree of change in the second formant frequency f2 is greater than that in the normal state. The degree of change of the second formant frequency f2 is represented by the fluctuation range, variance, etc. of the second formant frequency f2.

  A language spoken by a person is composed of a plurality of phonemes such as vowels and consonants. In particular, a person whose cognitive function is lowered is a vowel / a /, / i /, / u /, / e /, / o /. , The fluctuation range of the second formant frequency f2 tends to be large. Furthermore, among the vowels / a /, / i /, / u /, / e /, / o /, when a person with reduced cognitive function is uttering the vowel / a /, the second formant frequency f2 The fluctuation range tends to be large. It is not only the second formant frequency f2 but also other forms such as the first formant frequency f1 and the third formant frequency f3 that are affected when a person whose cognitive function is reduced is uttering the vowel / a /. The formant frequency is also affected.

  FIG. 3A shows the distribution of the formants of speech produced by people with reduced cognitive function. FIG. 3B shows a distribution of a formant of a voice uttered by a healthy person who has normal cognitive function. 3A and 3B, Δ represents the first formant frequency f1, ● represents the second formant frequency f2, and x represents the third formant frequency f3.

  In FIG. 3A, Wa is a period in which a person with reduced cognitive function emits a vowel / a /, and H21 is a fluctuation range of the second formant frequency f2 of the vowel / a / emitted by a person with reduced cognitive function. In FIG. 3B, let Wb be the period during which a healthy person continuously produces vowels / a /, and let H22 be the fluctuation range of the second formant frequency f2 of the vowels / a / emitted by the healthy person. 3A and 3B, the fluctuation range H21 of the second formant frequency f2 of the voice uttered by the person with reduced cognitive function is clearly compared with the fluctuation range H22 of the second formant frequency f2 of the voice uttered by the healthy person. You can see that it ’s big.

  In addition, each fluctuation range of the first formant frequency f1 and the third formant frequency f3 of the sound emitted by the person with the reduced cognitive function, and each of the first formant frequency f1 and the third formant frequency f3 of the sound emitted by the healthy person There is no significant difference between the fluctuation range.

  That is, the difference that appears between the formant of a person with reduced cognitive function and the formant of a healthy person is larger at the second formant frequency f2 than at the first formant frequency f1 and the third formant frequency f3. Therefore, it is preferable to use the feature quantity of the second formant frequency f2 for evaluation of the cognitive function.

  Therefore, the cognitive function evaluation apparatus 1 evaluates the cognitive function based on the above-described knowledge. FIG. 4 is a flowchart showing the cognitive function evaluation process of the cognitive function evaluation apparatus 1.

  The acquisition unit 11 acquires analog audio data from the microphone 21 (S1). The acquisition unit 11 monitors the audio data input from the microphone 21. When the effective value of the instantaneous sound pressure is equal to or greater than a predetermined value, the acquisition unit 11 performs an acquisition process of the audio data. Start. Note that the acquisition unit 11 may start the audio data acquisition process when the sound collection enabled state continues for a predetermined time or more. Moreover, the acquisition part 11 may acquire a subject's audio | voice in real time, or may acquire the recorded audio | voice.

  The acquisition unit 11 that has started the acquisition process of audio data converts analog audio data into digital audio data (A / D conversion function). Then, the acquisition unit 11 starts the audio data acquisition process, and then outputs the digital audio data to the extraction unit 12. That is, the acquisition unit 11 has an interface function of acquiring the subject's voice data and delivering the voice data to the subsequent stage. The audio data output from the acquisition unit 11 is a digital electric signal, and the audio data format is, for example, WAVE or MP3. The acquisition unit 11 may further convert voice data generated in a format such as WAVE or MP3 into voice data in CSV (Comma Separated Values) format and output the voice data. The acquisition unit 11 is set to, for example, 8 kHz as a sampling frequency for A / D conversion. However, the sampling frequency may be other than 8 kHz, and the specific value of the sampling frequency is not limited to a specific value.

  The microphone 21 may output digital audio data. In this case, the acquisition unit 11 receives digital audio data and passes the digital audio data to the extraction unit 12.

  The extraction unit 12 extracts audio data during a period in which the vowel / a / (specific phoneme) is uttered from the audio data received from the acquisition unit 11 (S2). For example, the extraction unit 12 has a voice recognition function and stores an acoustic model corresponding to the vowel / a / in advance. Then, the extraction unit 12 applies the acoustic model to the voice data received from the acquisition unit 11 and extracts the voice data during the period in which the vowel / a / is uttered. Hereinafter, the voice data extracted by the extraction unit 12 (voice data during a period when the vowel / a / is uttered) is referred to as target data.

  The formant analysis unit 13 obtains the second formant frequency f2 when the vowel / a / is uttered based on the target data (S3).

  Specifically, as shown in FIG. 5, the formant analysis unit 13 divides the target period Wc into a plurality of frames M1 when the period in which the target data is generated is the target period Wc. The frame M1 is set to a predetermined frame length. The generation period (frame period) T1 of the frame M1 is set to a predetermined period. In this embodiment, [frame period T1 <frame length of the frame M1] is set. In this case, the frame period T1 is preferably set to half the frame length of the frame M1. That is, in two adjacent frames M1, the subsequent period of the previous frame M1 overlaps the previous period of the subsequent frame M1.

  For example, the target period Wc is generally about 120 ms, the frame length of the frame M1 is set to 20 ms, and the frame period T1 is set to 10 ms. In this case, in the two frames M1 adjacent to each other, the 10 ms period in the latter half of the previous frame M1 overlaps the 10 ms period in the first half of the subsequent frame M1. When the target period Wc is 120 ms, twelve frames M1 are created for one target period Wc (see FIG. 5).

  The formant analysis unit 13 sets two or more frames M1 created near the center of the target period Wc among the plurality of frames M1 created within one target period Wc as analysis targets. The analysis target frame M1 is set as a target frame M11. In the present embodiment, in FIG. 5, among the 12 (first to twelfth) frames M1, the fourth to ninth six frames M1 are set as the target frame M11.

  The formant analysis unit 13 converts time-domain audio data into frequency-domain audio data for each of the target frames M11. The formant analysis unit 13 converts time domain audio data into frequency domain audio by DCT (Discrete Cosine Transform), FFT (Fast Fourier Transform), or WT (Wavelet Transform). Convert to data.

  Then, the formant analysis unit 13 obtains the second formant frequency f2 for each audio data of the target frame M11. That is, the formant analysis unit 13 can obtain the second formant frequency f2 corresponding to each of the six target frames M11 in one target period Wc.

  Next, the feature analysis unit 14 obtains a feature amount based on the second formant frequency f2 of each of the six target frames M11 in one target period Wc (S4). Specifically, the feature analysis unit 14 features a fluctuation range (difference between the maximum value and the minimum value of the second formant frequency f2) based on the second formant frequency f2 corresponding to each of the six target frames M11. Calculate as a quantity.

  FIG. 6A shows each data of six first formant frequencies f1 and six second formant frequencies f2 obtained from target data of a person whose cognitive function has deteriorated. FIG. 6B shows each data of six first formant frequencies f1 and six second formant frequencies f2 obtained from target data of a healthy person. 6A and 6B, Δ indicates the first formant frequency f1, and ● indicates the second formant frequency f2. It can be seen that the fluctuation range H23 of the second formant frequency f2 of the voice uttered by the person with the reduced cognitive function is clearly larger than the fluctuation width H24 of the second formant frequency f2 of the voice uttered by the healthy person.

  The feature analysis unit 14 obtains a difference between the maximum value of the second formant frequency f2 and the minimum value of the second formant frequency f2 among the six second formant frequencies f2. Then, the feature analysis unit 14 sets this difference as the fluctuation width of the second formant frequency f2 (the fluctuation width H23 in FIG. 6A and the fluctuation width H24 in FIG. 6B).

  The evaluation unit 15 evaluates the cognitive function of the subject based on the fluctuation range (feature amount) of the second formant frequency f2 (S5).

  Specifically, the evaluation unit 15 compares the fluctuation range of the second formant frequency f2 with a predetermined threshold value. If the fluctuation range of the second formant frequency f2 is equal to or greater than the threshold value, the evaluation unit 15 determines that there is a possibility that the cognitive function of the subject is degraded for the target period Wc. If the fluctuation range of the second formant frequency f2 is less than the threshold value, the evaluation unit 15 determines that the subject's cognitive function is normal for the target period Wc.

  Then, the evaluation unit 15 extracts a plurality of target periods Wc from the voice data of the target person, and performs the above determination process for each of the plurality of target periods Wc. And if the ratio of the frequency | count that the evaluation part 15 judged that cognitive function may have fallen with respect to all the frequency | counts of judgment is more than a predetermined value, the subject's cognitive function may have fallen. Assess that there is sex. In addition, the evaluation unit 15 evaluates that the cognitive function of the target person is normal if the ratio of the number of times it is determined that the cognitive function may be reduced is less than a predetermined value with respect to the total number of determinations. To do.

  The evaluation notification part 16 notifies the evaluation result of the cognitive function by the evaluation part 15 to a predetermined notification destination via the communication I / F 22 (S6).

  Furthermore, it is considered that the degree of reduction in cognitive function increases as the difference obtained by subtracting the threshold value from the fluctuation range of the second formant frequency f2 increases. Therefore, the evaluation unit 15 may select one of a plurality of cognitive function levels according to the degree of decrease in cognitive function. For example, the evaluation unit 15 evaluates the cognitive function level lower as the difference obtained by subtracting the threshold value from the fluctuation range of the second formant frequency f2 is larger. The evaluation notification unit 16 also includes this cognitive function level information in the evaluation result of the cognitive function.

  In addition, the above-mentioned threshold value is an experience value calculated | required by experiment, test, etc. of cognitive function evaluation. For example, the voice data of a person whose cognitive function has deteriorated and the voice data of a healthy person are collected in advance, and the collected voice data is analyzed in advance, whereby the threshold value is determined empirically.

  In FIG. 6A, the fluctuation ranges of the six second formant frequencies f2 obtained from the target data of the person whose cognitive function has deteriorated are H23. In FIG. 6B, the fluctuation range of the six second formant frequencies f2 obtained from the target data of the healthy person is H24. The magnitude relationship between the fluctuation range H23 of the second formant frequency f2 and the fluctuation range H24 of the second formant frequency f2 is H23> H24. The threshold used by the evaluation unit 15 is set to a value larger than the fluctuation range H23 and smaller than the fluctuation range H24. Therefore, the evaluation unit 15 can determine that the cognitive function of the subject may be reduced with respect to the fluctuation range H23 of the second formant frequency f2. Further, the evaluation unit 15 can determine that the cognitive function of the subject is normal with respect to the fluctuation range H24 of the second formant frequency f2.

  The cognitive function evaluation apparatus 1 is installed indoors such as a welfare facility, a day service center, a housing for elderly people with a service, and a housing of a target person. An indoor network NT1 is constructed indoors, and the communication I / F 22 is connected to the indoor network NT1 (see FIG. 1). The communication I / F 22 may have either a configuration connected by wire to the indoor network NT1 or a configuration connected wirelessly to the indoor network NT1. For example, when the communication I / F 22 is wirelessly connected to the indoor network NT1, a wireless router 4 is provided in the indoor network NT1, and the communication I / F 22 is connected to the indoor network NT1 via the wireless router 4. it can. The wireless communication specification used by the communication I / F 22 may be selected as appropriate from a wireless local area network (LAN), Bluetooth (registered trademark), and the like, and the wireless communication specification is not limited.

  Further, the information device 3 is connected to the indoor network NT1 (see FIG. 1). Therefore, the evaluation notification unit 16 can transmit the evaluation result of the cognitive function by the evaluation unit 15 to the indoor information device 3 via the communication I / F 22. As the information device 3, a smartphone, a tablet terminal, a personal computer, a mobile phone, or the like is used. The information device 3 is possessed by an indoor caregiver, family, or the like, and can acquire the evaluation result of the cognitive function of the target person and display it on the screen. Note that the information device 3 may be a dedicated terminal of the cognitive function evaluation system 10 or may indicate the evaluation result of the cognitive function using light, sound, or the like.

  Further, the indoor network NT1 is connected to the wide area network NT2 including the Internet via the wireless router 4 (see FIG. 1). The communication I / F 22 can communicate with the remote information device 5 from the indoor network NT1 via the wireless router 4, the wide area network NT2, the mobile communication network, and the like. Therefore, the evaluation notification unit 16 can transmit the evaluation result of the cognitive function by the evaluation unit 15 to the remote information device 5 via the communication I / F 22. As the information device 5, a smartphone, a tablet terminal, a personal computer, a mobile phone, or the like is used. The information device 5 is possessed by a remote family member or the like, and can acquire the evaluation result of the cognitive function of the target person and display it on the screen.

  The evaluation notification unit 16 performs, for example, a push notification that notifies the evaluation result of the cognitive function of the target person or creates an email. The device information or e-mail address of each of the information devices 3 and 5 is registered in the evaluation notification unit 16 in advance. The evaluation result to be notified includes a cognitive function level and the like.

  Moreover, when the cognitive function evaluation apparatus 1 makes the several subject person the evaluation object, it is preferable to change the notification destination of an evaluation result for every subject person. Therefore, the target identification unit 17 has a function of identifying the target person who has issued the voice data acquired by the acquisition unit 11. Specifically, the target identification unit 17 stores voice print data for each target person in advance, extracts the voice print data from the voice data acquired by the acquisition unit 11, and collates it with the voice print data for each target person. The target identifying unit 17 can identify the target person corresponding to the voice data based on the collation result.

  In the evaluation notification unit 16, information (device information, e-mail address, etc.) of a notification destination for each target person is registered in advance. The evaluation notification unit 16 selects a notification destination registered corresponding to the target person specified by the target identification unit 17, and notifies the selected notification destination of the evaluation result. Therefore, even if the cognitive function evaluation device 1 is a case where a plurality of subjects are to be evaluated, the cognitive function evaluation device 1 can identify the subject, so that only the notification destination related to the identified subject is evaluated. The result can be notified. The notification destination related to the specified target person is the information devices 3 and 5 possessed by the specified target person's family, a caregiver, and the like.

  Next, the modification 1 of the cognitive function evaluation apparatus 1 is demonstrated.

  FIG. 7 shows a configuration of the cognitive function evaluation apparatus 1 in the first modification. The cognitive function evaluation apparatus 1 in FIG. 7 further includes a model storage unit 18. And in the cognitive function evaluation apparatus 1 in the modification 1, as a feature quantity of the second formant frequency f2, six variation patterns of the second formant frequency f2 (showing how to move up and down), variation width, six The value of the second formant frequency f2 is used.

  The model storage unit 18 stores a learned statistical model that the evaluation unit 15 uses for the evaluation process of the cognitive function.

  The statistical model creation process will be described. First, voice data of a person whose cognitive function is lowered and voice data of a healthy person are collected in advance. Then, a machine learning computer executes machine learning using the collected voice data, and a statistical model representing a cognitive function evaluation processing algorithm is generated. This statistical model gradually evolves as the machine learning computer repeatedly performs machine learning, and the accuracy is improved. And a statistical model is written in the model memory | storage part 18 of the cognitive function evaluation apparatus 1 at any time, and is updated by the newest version. For example, a programming language such as Python is used as the machine learning tool, and an algorithm such as a random forest is used. However, the type is not limited. Note that the statistical model update processing for the model storage unit 18 is a wide-area communication network including an optical disk, a mode in which data is written to the model storage unit 18 from a portable recording medium such as a memory using a USB (Universal Serial Bus), and the Internet. And the like are written in the model storage unit 18 via

  The evaluation unit 15 applies the data of the six second formant frequencies f2 obtained in one target period Wc to the statistical model in the model storage unit 18. The evaluation unit 15 determines the subject's cognitive function for the target period Wc according to each feature amount such as the fluctuation pattern, fluctuation width, and each value of the six second formant frequencies f2 of the six second formant frequencies f2. Determine if there is a possibility of a decline.

  Then, the evaluation unit 15 extracts a plurality of target periods Wc from the voice data of the target person, and performs the above determination process for each of the plurality of target periods Wc. And if the ratio of the frequency | count that the evaluation part 15 judged that cognitive function may have fallen with respect to all the frequency | counts of judgment is more than a predetermined value, the subject's cognitive function may have fallen. Assess that there is sex. In addition, the evaluation unit 15 evaluates that the cognitive function of the target person is normal if the ratio of the number of times it is determined that the cognitive function may be reduced is less than a predetermined value with respect to the total number of determinations. To do.

  Therefore, the evaluation unit 15 can perform the cognitive function evaluation process using a plurality of feature amounts such as the variation pattern and variation width of the six second formant frequencies f2 and each value of the second formant frequency f2. . As a result, the evaluation unit 15 can improve the accuracy of the cognitive function evaluation.

  Next, Modification 2 of the cognitive function evaluation apparatus 1 will be described.

  FIG. 8 shows the configuration of the cognitive function evaluation apparatus 1 in the second modification. The cognitive function evaluation device 1 in FIG. 8 further includes a history storage unit 111, a trend analysis unit 112, and a trend notification unit 113.

  The history storage unit 111 sequentially stores evaluation results of the cognitive function by the evaluation unit 15 and stores a history of evaluation results (evaluation history). Since the target person corresponding to each evaluation result is specified by the target identifying unit 17, the history storage unit 111 stores the evaluation history in association with each target person.

  The trend analysis unit 112 can extract the temporal change of the evaluation result for each target person based on the evaluation history for each target person, and analyze the change tendency of the cognitive function for each target person. The trend analysis unit 112 executes this analysis processing regularly or when the number of evaluations of the subject reaches a predetermined number.

  For example, even if the evaluation result is normal, if the cognitive function level is gradually decreasing, the trend analysis unit 112 analyzes that the cognitive function tends to decrease. In this case, the trend notification unit 113 notifies the predetermined notification destination of the analysis result that the cognitive function tends to decrease via the communication I / F 22. At this time, it is preferable that the trend notification unit 113 notifies the notification destination of a message indicating that the cognitive function is reduced, a cognitive function improvement method, a method of suppressing the cognitive function decrease, and the like in addition to the analysis result.

  In addition, if the cognitive function level is in a short time, the trend analysis unit 112 analyzes that the cognitive function is rapidly deteriorating. In this case, the trend notification unit 113 notifies the predetermined notification destination via the communication I / F 22 of the analysis result that the cognitive function is rapidly deteriorating. At this time, it is preferable that the trend notification unit 113 notifies the analysis result as an emergency message.

  If the cognitive function level is increasing, the trend analysis unit 112 analyzes that the cognitive function is improved. In this case, the trend notification unit 113 notifies an analysis result that the cognitive function is improved to a predetermined notification destination via the communication I / F 22. At this time, it is preferable that the trend notification unit 113 adds a message for promoting further improvement and notifies it.

  The analysis result by the trend analysis unit 112 may be other than the analysis result described above. The trend analysis unit 112 may analyze a cognitive function change tendency based on a cognitive function level decrease amount, a cognitive function level decrease gradient, a cognitive function level increase amount, a cognitive function level increase gradient, or the like. preferable.

  Further, as the degree of change of the second formant frequency f2, the dispersion (variation) of the second formant frequency f2 may be used. In this case, the feature analysis unit 14 obtains the variance of the second formant frequency f2 corresponding to each of the six target frames M11 as a feature amount.

  In general, when the cognitive function is lowered, the maximum value, minimum value, average value, and median value of a specific formant frequency (for example, the second formant frequency f2) tend to change from normal. Therefore, the feature analysis unit 14 may obtain the maximum value, minimum value, average value, or median value of the frequency of a specific formant as the feature amount.

  For example, the feature analysis unit 14 obtains the maximum value, the minimum value, the average value, or the median value from the second formant frequencies f2 corresponding to each of the six target frames M11. The evaluation unit 15 can perform a cognitive function evaluation process by comparing the maximum value, the minimum value, the average value, or the median value of the second formant frequency f2 with a threshold value.

  The cognitive function evaluation system 10 including the cognitive function evaluation apparatus 1 may be housed inside a decorative product such as a pendant, for example. In this case, the caregiver or family member of the subject wears this decoration. Therefore, when the caregiver or family talks with the target person, the voice of the target person is collected without making the target person aware of it, and the cognitive function evaluation apparatus 1 can evaluate the cognitive function.

  FIG. 9 shows a block configuration of the robot 6 having a function of cognitive function evaluation. The robot 6 is a communication robot and can interact with the target person.

  The robot 6 includes a cognitive function evaluation device 1, a microphone 21, a communication I / F 22, a dialogue function unit 23, and a speaker 24.

  The dialogue function unit 23 includes a voice recognition unit 231, a dialogue control unit 232, a voice synthesis unit 233, and a dialogue data storage unit 234.

  The dialogue data storage unit 234 stores voice data of various words in advance. The voice recognition unit 231 can recognize the content spoken by the target person by comparing the voice data collected by the microphone 21 with the voice data of the dialogue data storage unit 234. The dialogue control unit 232 derives a response for establishing a conversation for the content spoken by the subject. The dialogue control unit 232 reads out voice data corresponding to the derived response from the dialogue data storage unit 234 and delivers it to the voice synthesis unit 233. The voice synthesizer 233 synthesizes the received voice data and outputs the synthesized voice data from the speaker 24 as voice. In this way, the robot 6 can interact with the subject by recognizing the content spoken by the subject and outputting a response for establishing a conversation. Then, when the subject interacts with the robot 6, the subject's voice is collected without making the subject aware of it, and the cognitive function evaluation device 1 can evaluate the cognitive function.

  FIG. 10 shows a configuration in which the cognitive function evaluation apparatus 1 is combined with a nurse call system 7 such as a welfare facility, a day service center, a housing for elderly people with a service, a hospital or the like. The nurse call system 7 includes a parent terminal 71 used by a caregiver, a nurse, etc. for a call, a child terminal 72 used by the subject for a call, and a main device 73 that controls communication between the parent terminal 71 and the child terminal 72. . The cognitive function evaluation device 1 is electrically connected to the main device 73 and collects voice data of the subject when the subject is talking using the child terminal 72. In this case, when the target person makes a call using the child terminal 72 of the nurse call system 7, the target person's voice is collected without making the target person conscious, and the cognitive function evaluation apparatus 1 evaluates the cognitive function. It can be performed.

  Moreover, the cognitive function evaluation system 10 provided with the cognitive function evaluation apparatus 1 may be accommodated in a mobile terminal capable of making a call such as a smartphone or a tablet terminal. An application program (application) for cognitive function evaluation is installed in the mobile terminal, and a computer built in the mobile terminal executes this program. As the microphone 21 and the communication I / F 22, a microphone and a communication I / F provided in the portable terminal are used. As a result, the mobile terminal operates as the cognitive function evaluation system 10. And when a subject talks using a portable terminal, a subject's audio | voice is collected without making a subject conscious, and the cognitive function evaluation apparatus 1 can evaluate a cognitive function. In addition, by using the portable terminal owned by the subject, the cognitive function evaluation for the elderly living alone can be easily performed.

  Moreover, the cognitive function evaluation system 10 provided with the cognitive function evaluation apparatus 1 may be accommodated in daily necessities such as a thermometer and a thermometer, for example. In this case, the target person can easily perform cognitive function evaluation for the elderly living alone by using the daily necessities owned by the target person.

  As a cognitive function evaluation method, there are an evaluation method using a question and an answer, such as MMSE (Mini Mental State Examination), HDS-R (Revised Hasegawa Simple Intelligence Evaluation Scale) and the like. Therefore, the cognitive function evaluation device 1 may evaluate the cognitive function using the voice data of the voice uttered by the target person at the time of answering when executing the evaluation method based on the question and the answer.

  Moreover, the apparatus provided with the function of the cognitive function evaluation apparatus 1 is not limited to the above-described decorative product, robot, and everyday product. For example, the cognitive function evaluation apparatus 1 or the cognitive function evaluation system 10 can be configured by a personal computer. Moreover, the structure which comprises the cognitive function evaluation apparatus 1 and communication I / F22 with a personal computer, and connects the microphone 21 to a personal computer may be sufficient. Alternatively, a dedicated device having the function of the cognitive function evaluation device 1 or a dedicated device having the function of the cognitive function evaluation system 10 may be used.

  As described above, the cognitive function evaluation device 1 according to the first embodiment of the present invention includes the formant analysis unit 13, the feature analysis unit 14, and the evaluation unit 15. The formant analysis unit 13 receives target data representing the temporal variation of the instantaneous sound pressure of a specific phoneme included in the target person's voice over the target period Wc. Then, the formant analysis unit 13 divides the target period Wc for the target data into a plurality of frames M1, and sets at least a specific formant frequency generated in the sound for each of two or more target frames M11 of the plurality of frames M1. Ask. The feature analysis unit 14 obtains a feature amount for a specific formant frequency obtained for each target frame M11. The evaluation unit 15 evaluates the cognitive function of the subject based on the feature amount.

  Since the above-described cognitive function evaluation device 1 evaluates the cognitive function of the subject using the formant frequency generated in the speech of the subject, the cognitive function evaluation device 1 is hardly affected by noise included in the speech data and distortion of the speech data. Therefore, the cognitive function evaluation apparatus 1 can perform evaluation with excellent robustness against noise and distortion of voice data in the cognitive function evaluation process. Furthermore, since the cognitive function evaluation apparatus 1 uses a specific formant frequency, the cognitive function evaluation process can be performed with high accuracy. In addition, the cognitive function evaluation apparatus 1 does not require advance preparation and can obtain the evaluation result of the cognitive function in a short time.

  That is, the above-described cognitive function evaluation apparatus 1 can obtain a highly accurate evaluation result of the cognitive function in a short time.

  Conventionally, as an evaluation method of cognitive function, evaluation by voice analysis, evaluation by blood test, evaluation by question and answer, evaluation by olfactory function, evaluation by analysis of behavior pattern, and the like are used.

  In the evaluation method based on speech analysis, a plurality of types of prosodic feature quantities extracted from speech data are stored in advance as learning data, and the risk of cognitive dysfunction is calculated by using this learning model. However, in the conventional evaluation method based on speech analysis, it takes a long time to calculate the risk of cognitive impairment by using a learning model. Furthermore, in the conventional evaluation method based on speech analysis, it is necessary to collect speech data of a speaker in advance and generate learning data, which is troublesome.

  The evaluation method by blood test evaluates cognitive function by detecting a specific biomarker by blood test. However, the evaluation method using a blood test requires a long time for the evaluation, and it takes about two to three weeks after the blood test until an evaluation result is obtained.

  Question and answer evaluation methods include MMSE (Mini Mental State Examination) and HDS-R (Revised Hasegawa-style Simple Intelligence Evaluation Scale). The target person answers and responds to a predetermined question. Evaluate cognitive function based on content and response status. However, in the evaluation method based on the question and answer, it takes about 10 minutes to 1 hour until the evaluation result is obtained after the target person answers. Furthermore, the evaluation method based on the question and the answer is a fixed question in which the question is determined in advance, and the evaluation accuracy may be lowered by repeatedly performing the same subject.

  The evaluation method using the olfactory function evaluates the cognitive function by performing a test in which the subject smells the odor and answers the odor to the answer sheet. However, the evaluation method using the olfactory function takes a long time for the evaluation after the test, and it takes about two to three weeks until the evaluation result is obtained.

  Also, the evaluation method based on behavior pattern analysis takes a long time for evaluation.

  Thus, the conventional evaluation method for cognitive function requires a long time for evaluation. Furthermore, conventional methods for evaluating cognitive functions require blood test equipment, test instruments for olfactory function, and the like, and the test cannot be performed easily. Moreover, in the conventional evaluation method of cognitive function, when the same subject is repeatedly evaluated, the accuracy of the evaluation may not be guaranteed.

  On the other hand, the cognitive function evaluation apparatus 1 can obtain the evaluation result of the cognitive function in a short time by using the feature quantity of the formant frequency of the voice. Specifically, the cognitive function evaluation apparatus 1 can reduce the time required for the evaluation to about 1 second after acquiring the voice data. And the cognitive function evaluation apparatus 1 can perform evaluation in real time, collecting a subject's audio | voice.

  Moreover, the cognitive function evaluation apparatus 1 analyzes the formant frequency of the subject's voice and extracts a feature amount. However, it is difficult for the subject to consciously adjust their formant frequency. Moreover, even if the subject consciously adjusts his / her formant frequency, it is difficult to continue that state. Therefore, the cognitive function evaluation apparatus 1 can maintain the test accuracy without deteriorating the test accuracy even when the test is repeatedly performed on the same subject.

  Moreover, the cognitive function evaluation device 1 does not need to use standard sentence voice data as the target person's voice data, and collects the subject's daily conversation, telephone voice, etc. without making the target person conscious, Cognitive function can be evaluated.

  Moreover, in the cognitive function evaluation apparatus 1 of the 2nd aspect which concerns on embodiment, it is preferable in a 1st aspect that a specific formant is the 2nd formant F2. Then, the feature analysis unit 14 obtains a feature amount for the frequency f2 of the second formant.

  In general, when the cognitive function is lowered, the feature amount of the second formant frequency f2 tends to change from the normal time. Therefore, since the cognitive function evaluation apparatus 1 uses the second formant frequency f2 that depends on the cognitive function, the cognitive function evaluation process can be performed with high accuracy.

  In the cognitive function evaluation apparatus 1 of the third aspect according to the embodiment, in the first or second aspect, the specific phoneme is preferably the vowel phoneme / a /.

  In general, when a person with reduced cognitive function utters a vowel / a /, the feature quantity of the formant frequency tends to change from normal. Therefore, the cognitive function evaluation apparatus 1 can perform a highly accurate cognitive function evaluation process.

  In the cognitive function evaluation device 1 of the fourth aspect according to the embodiment, in any one of the first to third aspects, the feature analysis unit 14 calculates the frequency of the specific formant obtained for each target frame M11. It is preferable to obtain the degree of change as the feature amount.

  In general, when cognitive function is lowered, the degree of change in frequency (such as fluctuation width or variance) of a particular formant tends to change from normal. Therefore, the cognitive function evaluation apparatus 1 can perform a highly accurate cognitive function evaluation process.

  In the cognitive function evaluation apparatus 1 of the fifth aspect according to the embodiment, in the fourth aspect, the degree of change is a fluctuation range or dispersion of the frequency of a specific formant obtained for each target frame M11. It is preferable.

  In this case, the cognitive function evaluation apparatus 1 can perform highly accurate cognitive function evaluation processing using the fluctuation range or dispersion of the frequency of a specific formant.

  Moreover, in the cognitive function evaluation apparatus 1 according to the sixth aspect according to the embodiment, in the fourth or fifth aspect, the evaluation unit 15 increases the degree of change in the frequency of the specific formant as the cognitive function of the subject. It is preferable to evaluate that has fallen.

  In general, when the cognitive function is lowered, the degree of frequency change (such as fluctuation width or variance) of a specific formant tends to be larger than that in a normal state. Therefore, the cognitive function evaluation apparatus 1 can perform a highly accurate cognitive function evaluation process.

  In the cognitive function evaluation device 1 of the seventh aspect according to the embodiment, in any one of the first to third aspects, the feature analysis unit 14 uses the frequency of the specific formant obtained for each target frame M11. It is preferable to obtain the maximum value, the minimum value, the average value, or the median value as the feature amount.

  Generally, when cognitive function is reduced, the maximum value, minimum value, average value, or median value of a particular formant tends to change from normal. Therefore, the cognitive function evaluation apparatus 1 can perform a highly accurate cognitive function evaluation process.

  Moreover, in the cognitive function evaluation apparatus 1 according to the eighth aspect according to the embodiment, in any one of the first to seventh aspects, a model storage that stores a statistical model representing a relationship between a feature amount and a cognitive function of a target person. It is preferable to further include a portion 18. And the evaluation part 15 applies a feature-value to a statistical model, and evaluates a subject's cognitive function.

  In this case, the cognitive function evaluation apparatus 1 can further improve the accuracy of the cognitive function evaluation.

  Further, in the cognitive function evaluation device 1 according to the ninth aspect according to the embodiment, in any one of the first to eighth aspects, an acquisition unit that acquires sound data representing temporal variation of the instantaneous sound pressure of the target person's sound. 11 and an extraction unit 12 that extracts target data from the audio data.

  In this case, the cognitive function evaluation device 1 can evaluate the cognitive function by collecting the subject's daily conversation, telephone voice, and the like without making the subject aware of it.

  In the cognitive function evaluation device 1 according to the tenth aspect according to the embodiment, it is preferable that the history storage unit 111 and the trend analysis unit 112 are further provided in any of the first to ninth aspects. The history storage unit 111 stores an evaluation history that is a history of evaluation results by the evaluation unit 15. The trend analysis unit 112 obtains a change tendency of the cognitive function of the target person based on the evaluation history.

  In this case, the cognitive function evaluation apparatus 1 can grasp the change tendency of the cognitive function of the subject based on the change in the cognitive function in the medium term or the long term. Therefore, even if the evaluation result is normal, the cognitive function evaluation device 1 can grasp that the cognitive function is in a decreasing tendency if the cognitive function level is gradually decreasing.

As described above, the cognitive function evaluation method of the eleventh aspect according to the embodiment of the present invention includes the following steps.
The formant analysis unit 13 receives target data representing the temporal variation of the instantaneous sound pressure of a specific phoneme included in the target person's voice over the target period Wc. Then, the formant analysis unit 13 divides the target period Wc for the target data into a plurality of frames M1, and sets at least a specific formant frequency generated in the sound for each of two or more target frames M11 of the plurality of frames M1. Step for obtaining (S3).
The feature analysis unit 14 obtains a feature amount for a specific formant frequency obtained for each target frame M11 (S4).
The evaluation unit 15 evaluates the subject's cognitive function based on the feature amount (S5).

  Therefore, the above-described cognitive function evaluation method can obtain a highly accurate evaluation result of the cognitive function in a short time.

  Moreover, the cognitive function evaluation apparatus 1 in FIG. 1 includes an acquisition unit 11, an extraction unit 12, a formant analysis unit 13, a feature analysis unit 14, an evaluation unit 15, an evaluation notification unit 16, and an object identification unit 17. Is integrated. However, the cognitive function evaluation apparatus 1 may include the above-described units in a distributed manner and connect the units via a network or the like. In this case, the cognitive function evaluation apparatus 1 is realized by using, for example, cloud computing technology. Alternatively, the cognitive function evaluation device 1 may be configured by one or more server devices.

  Moreover, the cognitive function evaluation apparatus 1 is equipped with a computer composed of a microcomputer or the like, and each function of the cognitive function evaluation apparatus 1 is realized by the computer executing a program. The computer installed in the cognitive function evaluation apparatus 1 includes a processor and an interface that operate according to a program as main hardware configurations. This type of processor includes a DSP (Digital Signal Processor), a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), etc., and can realize the function of the cognitive function evaluation apparatus 1 by executing a program. If possible, the type does not matter.

  As a program providing form, a computer-readable ROM (Read Only Memory), a form stored in advance in a recording medium such as an optical disc, or the like is supplied to the recording medium via a wide area communication network including the Internet. There are forms.

  That is, the program according to the twelfth aspect of the present invention causes a computer to function as the cognitive function evaluation apparatus 1. The program causes the computer to function as the formant analysis unit 13, the feature analysis unit 14, and the evaluation unit 15. The formant analysis unit 13 receives target data representing the temporal variation of the instantaneous sound pressure of a specific phoneme included in the target person's voice over the target period Wc. Then, the formant analysis unit 13 divides the target period Wc for the target data into a plurality of frames M1, and sets at least a specific formant frequency generated in the sound for each of two or more target frames M11 of the plurality of frames M1. Ask. The feature analysis unit 14 obtains a feature amount for a specific formant frequency obtained for each target frame M11. The evaluation unit 15 evaluates the cognitive function of the subject based on the feature amount.

  Therefore, the program of the present embodiment can obtain a highly accurate evaluation result of the cognitive function in a short time.

  The above-described embodiment is an example of the present invention. For this reason, the present invention is not limited to the above-described embodiment, and various modifications can be made depending on the design and the like as long as the technical idea according to the present invention is not deviated from this embodiment. Of course, it is possible to change.

DESCRIPTION OF SYMBOLS 1 Cognitive function evaluation apparatus 10 Cognitive function evaluation system 11 Acquisition part 12 Extraction part 13 Formant analysis part 14 Feature analysis part 15 Evaluation part 16 Evaluation notification part 17 Object identification part 18 Model storage part 111 History storage part 112 Trend analysis part 113 Trend notification Part Wc Target period M1 frame M11 Target frame F2 Second formant f2 Second formant frequency

Claims (12)

Receives target data representing temporal variation of instantaneous sound pressure of a specific phoneme included in the target person's voice over a target period, and divides the target period into a plurality of frames for the target data, resulting in the voice A formant analysis unit for obtaining at least a frequency of a specific formant for each of two or more target frames of the plurality of frames;
A feature analysis unit for obtaining a feature amount for the frequency of the specific formant obtained for each target frame;
An evaluation unit that evaluates the cognitive function of the subject based on the feature amount. The specific formant is a second formant;
The cognitive function evaluation apparatus according to claim 1, wherein the feature analysis unit obtains the feature amount for the frequency of the second formant.   3. The cognitive function evaluation apparatus according to claim 1, wherein the specific phoneme is a vowel phoneme / a /.   The cognitive function according to any one of claims 1 to 3, wherein the feature analysis unit obtains the degree of change in the frequency of the specific formant obtained for each target frame as the feature amount. Evaluation device.   The cognitive function evaluation apparatus according to claim 4, wherein the degree of change is a fluctuation range or variance of the frequency of the specific formant obtained for each target frame.   The cognitive function evaluation apparatus according to claim 4, wherein the evaluation unit evaluates that the cognitive function of the subject is lowered as the degree of change in the frequency of the specific formant is larger.   The feature analysis unit obtains a maximum value, a minimum value, an average value, or a median value of the frequency of the specific formant obtained for each target frame as the feature amount. The cognitive function evaluation apparatus as described in any one of thru | or 3. A model storage unit that stores a statistical model representing the relationship between the feature quantity and the cognitive function of the subject;
The cognitive function evaluation apparatus according to any one of claims 1 to 7, wherein the evaluation unit evaluates the cognitive function of the subject by applying the feature amount to the statistical model. An acquisition unit for acquiring audio data representing temporal variation of instantaneous sound pressure of the subject's voice;
The cognitive function evaluation apparatus according to any one of claims 1 to 8, further comprising an extraction unit that extracts the target data from the voice data. A history storage unit that stores an evaluation history that is a history of evaluation results by the evaluation unit;
The cognitive function evaluation apparatus according to any one of claims 1 to 9, further comprising: a tendency analysis unit that obtains a change tendency of the cognitive function of the subject based on the evaluation history. The formant analysis unit receives target data representing a temporal variation in instantaneous sound pressure of a specific phoneme included in the target person's voice over a target period, and divides the target period into a plurality of frames for the target data. Determining a frequency of at least a specific formant occurring in the speech for each of two or more target frames of the plurality of frames;
A step of obtaining a feature amount for the frequency of the specific formant obtained for each target frame by a feature analysis unit;
An evaluation unit includes a step of evaluating the cognitive function of the subject based on the feature amount. A cognitive function evaluation method, comprising: A program for causing a computer to function as a cognitive function evaluation device,
Receiving the target data representing the time variation of the instantaneous sound pressure of a specific phoneme included in the voice of the target person over the target period, dividing the target period into a plurality of frames for the target data, A formant analysis unit for obtaining a frequency of at least a specific formant occurring in speech for each of two or more target frames of the plurality of frames;
A feature analysis unit for obtaining a feature amount for the frequency of the specific formant obtained for each target frame;
A program that functions as an evaluation unit that evaluates the cognitive function of the subject based on the feature amount.

Images