LEARNING MODEL GENERATION DEVICE FOR DETERMINING CHEWING SIDE AND METHOD AND PROGRAM THEREFOR, AND CHEWING SIDE DETERMINATION DEVICE AND METHOD AND PROGRAM THEREFOR
PROBLEM TO BE SOLVED: To provide a reliable technology for determining a user's chewing side on the basis of surface myoelectric potential of a masseter muscle.SOLUTION: A chewing side determination device acquires a first and second myoelectric waveforms derived from a left and right muscles related to a user's chewing motion, respectively； calculates a correlation coefficient of information extracted from the respective first and second myoelectric waveforms as a first feature amount； calculates a second feature amount from a power spectrum obtained by performing frequency analysis on the first myoelectric waveform； calculates a third feature amount from a power spectrum obtained by performing frequency analysis on the second myoelectric waveform； generates a learning model by associating the first, second, and third feature amounts with a plurality of labels for distinguishing the user's chewing side； and refers to the learning model and determines the user's chewing side on the basis of a first, second, and third feature amounts calculated from newly acquired myoelectric waveforms.SELECTED DRAWING: Figure 2COPYRIGHT: (C)2020,JPO&INPIT【課題】 咀嚼筋の表面筋電位に基づいてユーザの咀嚼側を判定する信頼性の高い技術を提供する。【解決手段】 ユーザの咀嚼動作に係る左右の筋肉にそれぞれ由来する第１および第２の筋電波形を取得し、前記第１および第２の筋電波形の各々から抽出される情報の相関係数を第１の特徴量として算出し、前記第１の筋電波形を周波数解析して得られるパワースペクトルから第２の特徴量を算出し、前記第２の筋電波形を周波数解析して得られるパワースペクトルから第３の特徴量を算出し、前記第１、第２および第３の特徴量と、前記ユーザの咀嚼側を区別するための複数のラベルとを対応付けて学習モデルを生成し、前記学習モデルを参照して、新たに取得された筋電波形から算出される第１、第２および第３の特徴量に基づいて、前記ユーザの咀嚼側を判定する。【選択図】図２
