EMG-based Abnormal Gait Detection and Recognition
Yao Guo, Raffaele Gravina, Xiao Gu, Giancarlo Fortino, Guang‐Zhong Yang
Abstract
The early detection of gait abnormalities plays a key role in medical applications, where most of the previous abnormal gait recognition methods rely on kinematic data captured with vision-based systems or wearable inertial sensors. This paper, conversely, puts forward the ambitious objective to employ multiple wearable Electromyography (EMG) sensors for gait abnormalities detection. Our proposed approach uses eight wireless EMG sensors attached with skin electrodes on four muscles (i.e., Tibialis Anterior, Peroneus Longus, Gas-trocnemius, and Rectus Femoris) per each leg to measure the muscle response during walking activity. In the recognition stage, both meta-features with SVM and Bidirectional Long Short-Term Machine (BiLSTM) are exploited for gait abnormalities recognition from raw EMG data, Discrete Wavelet Transform (DWT) coefficients, and the reconstructed EMG signals, respectively. Experimental results on our gait dataset demonstrate the efficacy of EMG-based abnormal gait detection and recognition.