Research on Gait Recognition from EMG and IMU Signals Using Multimodal Residual - LSTM Networks

Gait recognition plays a crucial role in biometric authentication, rehabilitation medicine, and human-computer interaction. Traditional methods fail to fully exploit the rich timevarying characteristics of electromyography (EMG) signals. To address this issue, this paper proposes a Multimodal Residual Long Short Term Memory (MR-LSTM) method. By introducing a residual connection structure, this method effectively mitigates the vanishing gradient problem that traditional LSTM encounters when handling long sequences, significantly enhancing the model's ability to capture long term dependent features. Specifically, the proposed MR-LSTM method processes the EMG data of human lower limbs, and adaptively fuses multimodal features through an attention mechanism, enabling precise identification of gait features. Experimental results demonstrate that compared with traditional identification methods such as Standard LSTM, RLSTM (Based on EMG signal and IMU signal) and MR-LSTM (EMG - IMU), the MR-LSTM method proposed in this paper exhibits superior performance of accuracy, precision, recall and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{F 1}$</tex>-score in <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{3}$</tex> gaits recognition tasks.