Gait Recognition Based on A-Mode Ultrasound and Inertial Sensor Fusion Systems

This paper proposes a lower limb gait analysis based on the fusion of type A ultrasound and inertial sensors, and develops a related interactive control system. It designs and conducts gait experiments to explore its performance in human gait analysis. The paper selects appropriate experimental equipment and develops corresponding human-machine interaction interfaces to achieve connection, control, data display, collection, and storage for both types of devices. To validate signal fusion recognition performance, this paper conducts experiments with three gait types: walking on flat ground, upstairs, and downstairs, collecting extensive gait data. It explores feature extraction, dimensionality reduction, fusion, and classification methods. Four data fusion schemes are employed: vector concatenation, weighted fusion, maximum value fusion, and tensor fusion. The fusion schemes achieve higher recognition accuracy than using ultrasound (0.883) or IMU (0.840) alone, with weighted fusion and maximum value fusion reaching accuracies of 0.940 and 0.933, respectively. Specifically, the recognition accuracies using weighted fusion and maximum value fusion are 0.940 and 0.933, respectively, indicating high accuracy and fast calculation speed. These research results demonstrate that fusion of ultrasound and inertial sensing signals can effectively enhance the performance of human.