Gait Recognition in Different Terrains with IMUs Based on Attention Mechanism Feature Fusion Method

Gait recognition is significant in the fields of disease diagnosis and rehabilitation training by studying the characteristics of human gait with different terrain. To address the problem that the transformation of different outdoor terrains can affect the gait of walkers, a gait recognition algorithm based on feature fusion with attention mechanism is proposed. First, the acceleration, angular velocity and angle information collected by the inertial measurement unit is used; then the acquired inertial gait data is divided into periods to obtain the period data of each step; then the features are extracted from the data, followed by the visualization of the one-dimensional data into two-dimensional images. A lightweight model is designed to combine convolutional neural network with attention mechanism, and a new attention mechanism-based feature fusion method is proposed in this paper for extracting features from multiple sensors and fusing them for gait recognition. The comparison experimental results show that the recognition accuracy of the model proposed in this paper can reach 89%, and it has good recognition effect on gait under different terrain.