Robust and breathable all-textile gait analysis platform based on LeNet convolutional neural networks and embroidery technique

Abstract:Repetitive Strain Injury (RSI) and its related Musculoskeletal Disorders (MSDs) symptoms not only bring path ological pains to people, but also limit their physical activities and work abilities. The pathological changes in the footprints and other gait features provide a new way for the real-time monitoring and nursing of the recovery degrees of MSDs symptoms. In this work, based on the conformable, breathable, and lightweight all-fabric pressure sensing material, a novel highly robust universal platform ATPSA-LeNet, consisting of the all-textile pressure sensors array (ATPSA) and LeNet convolutional neural networks, has been proposed. Standing pos tures and authentication of volunteers have been identified from their gait characteristicswith high accuracy. The ATPSA-LeNet platform could directly convert foot pressure values into input data for the deep learning networks through the ATPSA, which greatly reduces the artificial errors arose from the spatial arranging of the sensors array and image data processing. Besides, ATPSA is more seamless and comfortable due to its improved compactness and breathability. Failures of sensing units also did not significantly decrease the overall accuracy. The proposed ATPSA-LeNet platform would provide a great prospect for extracting the high-dimensional spatial information contained in human gait features in many fields, such as clinical medicine, authentication, and criminal investigation

Zhao MM, Xu H, Zhong WB, et al.   Robust and breathable all-textile gait analysis platform based on LeNet convolutional neural networks and embroidery technique[J]. Sensors and Actuators: A. Physical.