Complex Motion Detection Based on Channel State Information and LSTM-RNN

Abstract

With the development of smart devices, human motion detection has been widely used for applications like entertainment and healthcare. Existing RF signal-based systems mostly focus on detecting relative strenuous actions and classifying them by Machine Learning (ML) method, like Support Vector Machine (SVM) and Random Forest (RF). This paper proposes a system that can detect and classify arm motions by leveraging the $W$ iFi OFDM signal. Instead of widely used SVM, we choose the Long Short-Term Memory (LSTM) algorithm to classify data from Channel State Information (CSI). The preliminary result shows that our systems achieve an average accuracy of 96% with 5 states of arm movement.