Human joint angle estimation with an e-textile sensor

Abstract

We describe the results of human joint angle estimation with an e-textile-based stretch sensor. Joint angle estimation is necessary to predict the details of human posture and implement posture instruction to prevent caregiver's injury in health and medical care. In this study, we focused on the elbow angle. We installed e-textile sensors on an elbow support to maintain an ideal setting and a knitted shirt to simulate a daily-use setting. With the elbow support, the correlation coefficient and the rooted-mean-square error in the degree of arc were 0.99 and 5.73 for complete bends. With the knitted shirt, in the same way, the coefficients and errors were 0.92 and 16.52 for complete bends and 0.71 and 7.30 for a daily action sequence. The range of motion of the subject's elbow was about 132°. Thus, our proposed system showed the potentials to detect up to 11 elbow angle patterns with an ideal setting and 9 patterns with the daily-use settings.