Human Activity Recognition Using WiFi Signal Features and Efficient Residual Packet Attention Network

Abstract

WiFi signal features, particularly channel state information (CSI), have gained considerable attention in human activity recognition (HAR) due to their nonintrusive and privacy–friendly nature. However, CSI packets are often nonstationary and exhibit fluctuations across various human activities. In this letter, we propose an end-to-end deep neural network (DNN) called efficient residual packet attention network (ERPANet) to tackle these challenges. In the proposed framework, we introduce the multilayer residual module composed of an attention residual (AR) operation and a downsampling attention residual (DAR) operation to effectively capture spatial-temporal features of CSI packets. In addition, a self-attention mechanism is embedded within AR and DAR to emphasize the importance of interrelationship among these multiscale CSI packet features. The proposed ERPANet aims to encode both channel information and long-range dependencies of CSI packet features. Extensive experiments show that ERPANet outperforms state-of-the-art methods, achieving average accuracies of 99.4% and 99.6% on the university of toronto human activity recognition (UT-HAR) and nanyang technological university human activity recognition (NTU-HAR) datasets, respectively.