WiFi CSI Based Passive Human Activity Recognition Using BLSTM-TCN

With the rapid development of wireless communication technologies, the application of WiFi signals in human activity recognition (HAR) has attracted increasing attention. Traditional methods mainly rely on machine learning algorithms and require manual extraction of features derived from Channel State Information (CSI) in either the time or frequency domain. However, such features often fail to effectively capture the continuous temporal characteristics of human activities. To address this issue, this paper proposes a deep learning-based recognition approach using WiFi CSI, which builds an integrated model by fusing Bidirectional Long Short-Term Memory (BLSTM) and a Temporal Convolutional Network (TCN). The BLSTM module is employed to extract forward and backward temporal dependencies, while the TCN further captures both local and global deep temporal features. Experimental results demonstrate that, compared with several baseline methods, the proposed approach significantly improves recognition accuracy.