Information processing device, information processing method, information processing system and information processing program

An information processing program executable by a computer to perform processing, comprising:

• at least one memory and at least one processor coupled to the memory, wherein the processor is configured to:
• acquire voice data that is time series data of a voice uttered by a user;
• conduct a frequency analysis on the voice data, wherein the frequency analysis includes determining an intensity of a frequency component;
• generate a spectrogram image of the voice data, wherein the spectrogram image includes a pixel value corresponding to the intensity of the frequency component;
• extract a feature amount, which is a predetermined acoustic parameter, from the voice data, to input the generated spectrogram image to a learned neural network and extract a second feature amount which comprises values output from an intermediate layer of the learned neural network model;
• calculate a score, which expresses an extent to which the user has a psychological disorder or a neurological disorder, or a psychological disturbance symptom or a cognitive impairment symptom, on the basis of the first feature amount and the second feature amount and a calculation model that is set in advance and is for calculating, from the first feature amount, a score expressing an extent of a psychological disorder or a neurological disorder, or a psychological disturbance symptom or a cognitive impairment symptom;
• and
• infer whether or not the user has any disorder or symptom, in accordance with the score,
• wherein the learned model is a learned model that has been learned in advance by teacher data in which a spectrogram image for learning, and a correct answer label expressing a disorder or symptom of a user who uttered the voice data corresponding to the spectrogram image for learning, are associated with one another.

An information processing apparatus, comprising:

• at least one memory and at least one processor coupled to the memory, wherein the processor is configured to:
• acquire voice data that is time series data of a voice uttered by a user;
• conduct a frequency analysis on the voice data, wherein the frequency analysis includes determining an intensity of a frequency component;
• generate an image corresponding to the voice data wherein the image includes a pixel value corresponding to the intensity of the frequency component
• extract a first feature amount, which is a predetermined acoustic parameter, from the voice data;
• input the first feature amount into a first learned model that has been learned in advance and is for calculating, from the feature amount, a score expressing an extent of a psychological disorder or a neurological disorder, or a psychological disturbance symptom or a cognitive impairment symptom, and to calculate a first score expressing an extent to which the user has a psychological disorder or a neurological disorder, or a psychological disturbance symptom or a cognitive impairment symptom;
• and
• infer whether or not the user has any disorder or symptom, in accordance with the score,
• wherein the learned neural network model is a learned neural network model that has been learned in advance by teacher data in which a spectrogram image for learning, and a correct answer label expressing a disorder or symptom of a user who uttered the voice data corresponding to the spectrogram image for learning, are associated with one another.

An information processing method comprising:

• by a processor:
• acquiring voice data that is time series data of a voice uttered by a user;
• conducting a frequency analysis on the voice data, wherein the frequency analysis includes determining an intensity of a frequency component;
• generating a spectrogram image of the voice data, wherein the spectrogram image includes a pixel value corresponding to the intensity of the frequency component;
• extracting a feature amount, which is a predetermined acoustic parameter, from the voice data, to input the generated spectrogram image to a learned neural network model, and extracting a second feature amount, which comprises values output from an intermediate layer of the learned neural network model;
• calculating a first score, which expresses an extent to which the user has a psychological disorder or a neurological disorder, or a psychological disturbance symptom or a cognitive impairment symptom, on the basis of the first feature amount and the second feature amount and a calculation model that is set in advance and is for calculating, from the first feature amount and the second feature amount, a score expressing an extent of a psychological disorder or a neurological disorder, or a psychological disturbance symptom or a cognitive impairment symptom;
• and
• inferring whether or not the user has any disorder or symptom, in accordance with the composite score,
• wherein the learned neural network model is a learned neural network model that has been learned in advance by teacher data in which a spectrogram image for learning, and a correct answer label expressing a disorder or symptom of a user who uttered the voice data corresponding to the spectrogram image for learning, are associated with one another.

A non-transitory storage medium storing an information processing program executable by a computer to perform processing, the processing comprising:

• acquiring voice data that is time series data of a voice uttered by a user;
• conducting a frequency analysis on the voice data, wherein the frequency analysis includes determining an intensity of a frequency component
• generating a spectrogram image of the voice data, wherein the spectrogram image includes a pixel value corresponding to the intensity of the frequency component;
• extracting a feature amount, which is a predetermined acoustic parameter, from the voice data, to input the generated spectrogram image to a learned neural network model, and extracting a second feature amount, which comprises values output from an intermediate layer of the learned neural network model;
• calculating a first score, which expresses an extent to which the user has a psychological disorder or a neurological disorder, or a psychological disturbance symptom or a cognitive impairment symptom, on the basis of the first feature amount and the second feature amount and a calculation model that is set in advance and is for calculating, from the first feature amount and second feature amount, a score expressing an extent of a psychological disorder or a neurological disorder, or a psychological disturbance symptom or a cognitive impairment symptom;
• and
• inferring whether or not the user has any disorder or symptom, in accordance with the composite score,
• wherein the learned neural network model is a learned neural network model that has been learned in advance by teacher data in which a spectrogram image for learning, and a correct answer label expressing a disorder or symptom of a user who uttered the voice data corresponding to the spectrogram image for learning, are associated with one another.