運動・生理信号を用いた超重症児の気持ち認識システムの模擬実験による検討

抄録

SMID-MCDG are children with severe motor and intellectual disabilities medical care dependent group. Most of them have difficulty expressing themselves using their voice and face. Therefore, those who take care of them pay attention to their motion and physiological signals to understand their emotion. However, it cost much time to become to be able to understand their emotion because their motion signals are unclear and physiological signals change for various reasons. To solve this problem, I tried to develop emotion recognition system for SMID-MCDG using motion and physiological signals measured by wearable sensor. This system cannot speak for them but might be clues for communication. As a first step, I considered the methods of emotion recognition based on the data collected by simulation experiment. The simulation experiment were conducted in supine posture to simulate that SMID-MCDG is bedridden. In the experiment, pleasant, unpleasant or neutral were evaluated per minute. In the emotion recognition, the number of motions for one minute and heart rate variability are chosen as features and decision tree is used as classification method. As the result, proposed method showed about 64.5% of accuracy rate for the data of simulation experiment.