Research On Device-Free Multi-Target Tracking With Wireless Sensor Networks

In the communication and information technology field, target localization and tracking has always been a popular research subject, which has a great potential and a bright prospect for commercial, civil and military application. Most of the existing wireless sensor network (WSN) based tracking methods require targets to carry transceiver devices, by which the application range is greatly limited. Device-free tracking methods do not require targets to carry any electronic device so that it substantially extend the range of practical application. However, existing device-free tracking methods have two major defects. First, existing measurement models can be unreliable in an indoor environment. Second, no tracking method has been proposed for tracking a time-varying number of targets. In this paper, a method is proposed to solve the two defects and accomplish multi-target localization and tracking in an indoor environment.In this paper, we first propose a received signal strength (RSS) histogram-based measurement model, which includes an on-line learning algorithm and a kernel density estimation (KDE)-based likelihood function computation approach. Second, a novel multi-target tracking method is proposed, which incorporates two major contributions:1. a multi-Bernoulli approximation-based filtering algorithm and its particle implementation;2. a likelihood function computation approach based on combination of tracking and detection results. Experimental results and error analysis are presented and discussed in detail, which demonstrates that the proposed tracking method achieves high performance both in target number estimation and target position estimation.