A CSI Based Localization and Identification Recognition Algorithm Using Multi-task Learning and Deep Residual Shrinkage Network

With the increasing requirements of wireless sensing technology, device-free based wireless sensing technique has received much attentions. In this paper, the channel state information (CSI) obtained from Wi-Fi signal is proposed for localization and identification using multi-task learning and deep residual shrinkage network (DRSN). First, the Hampel filter and wavelet filter are used to remove outliers and mitigate noise, respectively. Then, the hard parameter sharing based multi-task learning framework is used for localization and identification recognition training. Specially, the DRSN is used for feature extraction in the shared hidden layer. The loss function is defined as the weight sum of the loss function of each task. The experimental results show that the accuracy of localization and identification of the proposed algorithm are 89.2% and 92.3%, respectively. It can achieve high precision joint task identification of location and identity.