Research Article
Human Motion Recognition Based on Wi-Fi Imaging
@INPROCEEDINGS{10.1007/978-3-030-92635-9_35, author={Liangliang Lin and Kun Zhao and Xiaoyu Ma and Wei Xi and Chen Yang and Hui He and Jizhong Zhao}, title={Human Motion Recognition Based on Wi-Fi Imaging}, proceedings={Collaborative Computing: Networking, Applications and Worksharing. 17th EAI International Conference, CollaborateCom 2021, Virtual Event, October 16-18, 2021, Proceedings, Part I}, proceedings_a={COLLABORATECOM}, year={2022}, month={1}, keywords={Motion recognition Wi-Fi image SVM}, doi={10.1007/978-3-030-92635-9_35} }- Liangliang Lin
Kun Zhao
Xiaoyu Ma
Wei Xi
Chen Yang
Hui He
Jizhong Zhao
Year: 2022
Human Motion Recognition Based on Wi-Fi Imaging
COLLABORATECOM
Springer
DOI: 10.1007/978-3-030-92635-9_35
Abstract
The current wireless sensing technology has some problems, such as low resolution caused by narrow signal bandwidth, poor environmental adaptability caused by multipath effect and so on. To solve the above problems, this paper provides a new perception idea, Wi-Fi imaging human actions, and then using image processing method for action recognition. In the Wi-Fi imaging part, according to the different spatial angles of different parts of the body trunk such as head, chest and legs relative to the receiving end, this paper processes the human body reflection signal received by the receiving end, obtains the signal strength corresponding to each azimuth signal in the space, and generates the human body heat map. In the stage of action recognition, firstly, the background interference is removed. According to the characteristics of imaging and action in this paper, a continuous action segmentation method is proposed. The image action features are obtained through intensive sampling. Finally, the SVM is optimized by genetic algorithm to improve the accuracy of action classification under different conditions. Through the analysis of experimental results, the method proposed in this paper can produce high-precision imaging of human body under practical application conditions. The recognition accuracy of different actions in the experiment is more than 90%.
