Research on Human Activity Recognition Technology under the Condition of Through-the-wall

Abstract

Wi-Fi-based human activity recognition is playing a critical role in wireless sensing. However, the existing through-wall human activity recognition method does not fully analyze the influence of the wall on the signal, which results in poor robustness of the Wi-Fi indoor human activity recognition system. In order to solve this problem, this paper proposes a Wi-Fi based activity recognition algorithm under through-the-wall scenarios. First, the distribution of Wi-Fi signals in the presence of wall barriers is analyzed according to the Wi-Fi signal model. Then, according to the distribution characteristics of different Wi-Fi signals, the principal component analysis (PCA) algorithm is used to reconstruct the signal to complete the de-noising processing of the Wi-Fi signal. Finally, feature extraction and feature classification in the time-frequency domain is performed to complete the human activity recognition. It is worth mentioning that in terms of feature extraction, we innovatively use the empirical mode decomposition (EMD) algorithm to extract the difference in time series of similar actions. Experimental results show that the system achieves an average accuracy of 95.82 percent in through-the-wall scenarios.