Dynamic Associate Domain Adaptation for Human Activity Recognition Using WiFi Signals

In this paper, a semi-supervised transfer learning with dynamic associate domain adaptation is proposed for human activity recognition by using the channel state information (CSI) of the WiFi signal. We propose a dynamic associate domain adaptation (DADA), by modifying the existing associate domain adaptation algorithm, while the target domain can dynamically provide a different ratio of labelled data set/unlabelled data set. The advantage of DADA is that it provides a dynamic strategy to eliminate different effects under the different environments. We designed an attention-based DenseNet model (AD) as our training network, so our proposed scheme is simplified as DADAAD scheme. The experimental results illustrate that the accuracy of human activity recognition of the DADA-AD scheme is 97.4%. It also shows that DADA-AD has advantages over existing semi-supervised learning schemes.