Learning model-generating apparatus, method, and program for assessing favored chewing side as well as determination device, method, and program for determining favored chewing side

Patent metadata

Field	Value
Publication number	US-11596336-B2
Application number	US-201917059039-A
Country	US
Kind code	B2
Filing date	May 22, 2019
Priority date	May 30, 2018
Publication date	Mar 7, 2023
Grant date	Mar 7, 2023

Abstract

A reliable technology for determining the masticatory side of the user is provided. First and second electromyographic waveforms respectively originating from left and right muscles related to masticatory actions of a user are acquired; a coefficient of correlation between pieces of information respectively extracted from the first and the second electromyographic waveforms is calculated as a first feature value; a second feature value is calculated from a power spectrum obtained by performing frequency analysis on the first electromyographic waveform; a third feature value is calculated from a power spectrum obtained by performing frequency analysis on the second electromyographic waveform; a learning model is generated by associating the first, second, and third feature values with a plurality of labels; and the masticatory side of the user is determined based on first, second, and third feature values calculated from a newly acquired electromyographic waveform and the learning model.

First claim

The invention claimed is:

1. A masticatory side determination apparatus that determines a masticatory side of a user, the masticatory side determination apparatus comprising:
a processor; and
a storage medium having computer program instructions stored thereon, when executed by the processor, perform to:
acquire first and second electromyographic waveforms respectively originating from left and right muscles related to masticatory actions of the user;
calculate a first feature value as a coefficient of correlation between pieces of information respectively extracted from the first and the second electromyographic waveforms, the coefficient of correlation being a cross-coefficient of correlation between absolute values of potentials determined based on root mean squared (RMS) processing of the first and second electromyographic waveforms;
calculate a second feature value from a power spectrum obtained by performing frequency analysis on the first electromyographic waveform, the second feature value being a median power frequency from the power spectrum of the first electromyographic waveform;
calculate a third feature value from a power spectrum obtained by performing frequency analysis on the second electromyographic waveform, the third feature value being a median power frequency from the power spectrum of the second electromyographic waveform; and
generate a learning model based on the first, second, and third feature values; and
determine the masticatory side of the user using the learning model, based on the first, second, and third feature values.

2. The masticatory side determination apparatus according to claim 1 , wherein the computer program instructions further perform to determine whether the first, second, and third feature values are normal values or abnormal values for each predetermined unit time range, based on the first, the second, and the third feature values, and performs masticatory side determination only for a unit time range in which the feature values are determined as normal values.

3. The masticatory side determination apparatus according to claim 2 , wherein the computer program instructions use an unsupervised learning model to perform abnormal value determination.

4. A non-transitory computer-readable medium having computer-executable instructions that, upon execution of the instructions by a processor of a computer, cause the computer to function as the masticatory side determination apparatus of claim 1 .

5. A learning model generation method that is carried out by a learning model generation apparatus, comprising:
acquiring first and second electromyographic waveforms respectively originating from left and right muscles related to masticatory actions of a user;
calculating a first feature value as a coefficient of correlation between pieces of information respectively extracted from the first and the second electromyographic waveforms, the coefficient of correlation being a cross-coefficient of correlation between absolute values of potentials determined based on root mean squared (RMS) processing of the first and second electromyographic waveforms;
calculating a second feature value for learning, from a power spectrum obtained by performing frequency analysis on the first electromyographic waveform, the second feature value being a median power frequency from the power spectrum of the first electromyographic waveform;
calculating a third feature value for learning, from a power spectrum obtained by performing frequency analysis on the second electromyographic waveform, the third feature value being a median power frequency from the power spectrum of the second electromyographic waveform;
generating a learning model based on the first, second, and third feature values and a plurality of labels for specifying a masticatory side of the user; and
determining the masticatory side of the user using the learning model, based on the first, second, and third feature values.

Assignees

• Nippon Telegraph & Telephone

Inventors

• Niijima Arinobu
• Isezaki Takashi
• Aoki Ryosuke
• Watanabe Tomoki
• Yamada Tomohiro

Classifications

• A61B5/7207

  of noise induced by motion artifacts · CPC title

• A61B5/397

  Analysis of electromyograms · CPC title

• A61B5/7264

  Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems · CPC title

• A61B5/228 Primary

  of masticatory organs, e.g. detecting dental force · CPC title

• A61B5/389

  Electromyography [EMG] · CPC title