Device-Free Human Activity Recognition Based on Channel Statement Information

The human activity recognition system has been researched in various technical fields for decades. Nowadays, as WiFi is widely used in homes, using WiFi devices for human activity recognition has become a better choice. In recent years, more and more researchers have begun to use channel state information (CSI) to realize human activity recognition. The CSI features with fine-grained information can show the impact of human activity on the channel. But in the existing CSI-based human activity recognition system, there is an issue. Due to the high proportion of error and noise in the phase information in CSI, the information in CSI is not fully utilized during the processing of CSI data. In this paper, we propose a method for extracting phase information in CSI, so that we can completely extract the effective information in CSI as the input feature of the classifier. Then, we use k-means for feature extraction of main feature data. Finally, we use support vector machines (SVM) to learn features and conduct activity recognition. We evaluated the system performance, and the experimental results show that our system has good performance.