Deep learning-based discriminant model for wearable sensing gait pattern

Abstract

<p>In order to effectively improve the accuracy of identifying the gait pattern of wearable sensing data, this paper proposes a new model for deep learning gait mode discrimination that integrates convolutional neural network and long short-term memory neural network, which makes full use of the convolutional neural network to obtain the most local spatial characteristics of data and the long short-term memory neural network to obtain the inherent characteristics of the data, and effectively excavates the hidden high-dimensional, nonlinear, time-space gait characteristics of random wearable sensing timing gait data that are closely related to gait pattern changes, to improve the classification performance of gait mode. The effectiveness of the proposed model in this paper is evaluated using the HAR dataset from University of California UCI database. The experiment results showed that the proposed model in this paper can effectively obtain the time-space gait characteristics embedded in the wearable sensor gait data, and the classification accuracy can reach 91.45%, the precision rate 91.54%, and the recall rate 91.53%, and the classification performance is significantly better than that of the traditional machine learning model, which provides a new solution for accurately identifying the gait mode of wearable sensor data.</p>