IMU-Based Gait Phase Recognition for Stroke Survivors: Preliminary Results

In this paper, we present an Inertial Measurement Unit (IMU) based gait phase detection system for stroke survivors. The system consists of two IMUs tied to the thigh and shank respectively, for collecting acceleration and angular velocity during walking. Features are extracted using a 150ms sliding window and processed by a quadratic discriminant analysis classifier. Three stroke survivors were recruited with varying degrees of walking disability to test our system, and the experimental environment was level walking at preferred speeds. Experimental results show that the IMU-based gait phase detection system can accurately identify the swing phase and the stance phase, with a recognition accuracy higher than 97 %. Also, we figure out that the recognition results of utilizing one IMU alone is almost equal to the results of using two IMUs together. This study provides an idea for further research on wearable rehabilitation robots for stroke survivors.