Research on Human Gait Recognition Based on the Improved BECNN-BILSTM

This study proposes a feature optimization strategy based on mutual information to identify the 10 most critical plantar pressure sensors for gait recognition. By reducing data redundancy, this approach enhances both the efficiency and generalization capability of the model. An advanced BECNN-BILSTM algorithm is developed to focus on key channel features, improving feature representation while suppressing irrelevant information. This design facilitates the accurate recognition of eight distinct lower-limb gait patterns. Experimental results demonstrate that the BECNN-BILSTM model significantly outperforms CNN, LSTM, and CNN-BILSTM models in recognition performance. This work offers a practical and efficient solution for real-time gait monitoring and fall prevention posture detection in wearable device applications.