Human Activity Recognition Using Deep Learning Networks with Enhanced Channel State Information

Abstract

Channel State Information (CSI) is widely used for device free human activity recognition. Feature extraction remains as one of the most challenging tasks in a dynamic and complex environment. In this paper, we propose a human activity recognition scheme using Deep Learning Networks with enhanced Channel State information (DLN-eCSI). We develop a CSI feature enhancement scheme (CFES), including two modules of background reduction and correlation feature enhancement, for preprocessing the data input to the DLN. After cleaning and compressing the signals using CFES, we apply the recurrent neural networking (RNN) to automatically extract deeper features and then the softmax regression algorithm for activity classification. Extensive experiments are conducted to validate the effectiveness of the proposed scheme.