RESEARCH ON MOTION RECOGNITION BASED ON CHANNEL STATE INFORMATION

Aiming at the problems that Fourier transform cannot process non-stationary signals and wavelet transform cannot identify the frequency changes in windows, a data processing method based on adaptive variational modal decomposition is proposed. The adaptiveness of the empirical modal decomposition method is used to determine the signal decomposition layers of the variational mode, and the channel state information (CSI) energy signal is decomposed. Based on the more fine-grained CSI signal, according to the characteristics of CSI when no human performs the actions indoors, the direction of arrival on the receiving antenna is monitored, and the indoor human action detection is realized. The effectiveness of the method in the laboratory environment is verified, and the influence of different parameters on action recognition is discussed. The experimental results show that the accuracy of the proposed method is about 93.3%.