Estimation of mind state with body movements observed using video cameras

摘要

This paper proposes a method for estimating the state of mind on observation of the body motion from the front and the side of a subject using 2 cameras. There are many researches that studied about the estimation of a psychological condition using facial features. For example, there are studies about facial expressions, eyes' motion and the facial thermal images using a specific camera. Observation of the movement of a body is easier and takes lower cost than one of the facial images. We especially focused on the situation when people feel bored and pleasant. We used the correlation analysis method to find which body motion correlates to the states of a mind. To estimate the state of a mind the authors have focused on the average amount of body motion, and the power (amplitude) of motion in each frequency range. Our experiment shows that when subject feels bored, stretching body up and down, and movements between left and right tend to increase that can be observed from both cameras. Especially, the result of both cameras has shown that the amplitude of stretching body up and down correlates to the state of mind in frequency range from 2.8Hz to 4.0Hz. The varieties of the subjects show many possible dependencies between body motions and the state of a subject's mind.