The Recognition of Human Activities Under UWB Communication

This paper presents a novel human activity recognition method that uses UWB signals to enable a low clutter outdoor environment sensing and recognition of human activities, which can transmit the information and identify human activities simultaneously. Since UWB signals do not require line-of-sight and have very good ability of penetration, the proposed method can enable a low clutter outdoor environment human activities recognition using the UWB signals in wireless communication. Further, it achieves this goal for a through-targets scenario and without requiring seeing devices (e.g., camera, radar). We evaluate the proposed method using UWB signals in a playground, with eight human subjects performing eight different activities. The type of the human activities performed between the transmitter and receiver of UWB communication system can have significant effects on the shape of the received signal waveform. From these time-varying signals, we extract features that are representative of the activities types based on 1-D diagonal slice of fourth-order cumulant within a time window. Then, we use support vector machine (SVM) to realize the human activities identification. Our results show that proposed method can identify and classify a set of eight activities with an average accuracy of 99.2 %.