An improved model for contactless gait recognition

Abstract

For Wi-Fi signal perception, aiming at problems of insufficient acquisition of feature and low recognition accuracy in multi-person scene in gait recognition, we propose a new gait recognition model WiMGNet based on energy distribution map(EDM). Depending on the channel response information impact factor analysis, WiMGNet uses the mechanism EDM to reconstruct the raw data effectively, so that it can contain more gait features. Furthermore, WiMGNet introduces EDM into neural network model, which obtains a high accuracy in gait recognition in multi-person scene. Compared to current gait recognition models, WiMGNet significantly improves the ability of feature acquisition and recognition accuracy. The experimental results show that WiMGNet has a recognition accuracy of 98.8% in 30-person scene experiment in indoor environment, which has obvious advantages compared to other similar models.