A cycling wheelchair (CWC) is a rehabilitation tool for hemiplegic patients. In previous studies, we proposed some indices to evaluate the motor function of the paralyzed leg of the patient driving a CWC. In this study, we focused on the involuntary leg abduction while rotating a CWC pedal, and proposed a new evaluation index based on the abduction motion. The leg abduction can be associated with motor dysfunction because it is due to the contracted muscles of the paralyzed leg. The proposed index requires signals from the gyro and accelerometer sensors attached to the shin of a CWC user. Based on these signals, the abduction angle and the knee position were estimated using a Kalman filter and a three-dimensional model of the leg, respectively. The proposed index is defined as the ratio between variances of the first and second main components of the leg motion trajectory. The main components were calculated by the principle component analysis. Experimental results obtained by employing healthy subjects, who imitated the leg abduction, showed that the proposed method is useful for evaluating the differences in the leg abduction patterns.
