Discrimination of Normal Side Gait Changes in Gait Asymmetry Using a Multi-Kernel Convolution Model

We propose a novel approach for discriminating the effects of gait asymmetry on unilateral normal lower limb gait variations, based on a multi-kernel convolutional deep learning model. This method considers the identification of gait asymmetry effects on unilateral normal lower limb gait variations as a classification problem. It effectively utilizes the multi-scale dynamic receptive field features of the multi-kernel convolutional deep learning model to extract significant dynamic variation features from the fused gait data of the unilateral lower limb joints. This approach enhances the generalization performance of unilateral lower limb gait classification and accurately identifies the changes in gait caused by gait asymmetry effects on the uni-lateral normal lower limb. total of 21 healthy participants were recruited to simulate different conditions of gait asymmetry by wearing orthopedic boots, height-increasing shoes, and sandbags on their unilateral lower limbs. The effectiveness of the proposed method was evaluated by utilizing the DELSYS wearable wireless surface electromyography (sEMG) collection system to gather sensor data from the hip, knee, and ankle joints of both lower limbs. The results showed that the proposed method achieved an accuracy rate of 99% in accurately identifying gait asymmetry caused by wearing orthopedic boots, height-increasing shoes, and sandbags, with better generalization performance compared to the comparison models. Moreover, the proposed method achieved precision rates of over 90% and recall rates of over 86% in accurately identifying the corresponding six categories of gait differences caused by gait asymmetry in different conditions. This demonstrates that the multi-kernel convolutional deep learning model can effectively extract dynamic variation features of unilat-eral normal lower limb gait caused by gait asymmetry, facilitating early identification of gait impairments in the unilateral normal lower limb.