Adaptive Filtering Methods for RSSI Signals in a Device-Free Human Detection and Tracking System

• Published in: IEEE systems journal Vol. 13; no. 3; pp. 2998 - 3009
• Main Authors: Booranawong, Apidet, Jindapetch, Nattha, Saito, Hiroshi
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.09.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; adaptive filtering methods; Adaptive filters; Base stations; Biomedical monitoring; Complexity; device-free detection and tracking; Error detection; Human motion; human movements; Indoor environments; Monitoring; Motion perception; Pattern analysis; Radio transmitters; Received signal strength indicator; Receivers; RSSI; Signal strength; Tracking; Tracking devices; Tracking errors; Tracking systems
• ISSN: 1932-8184; 1937-9234
• DOI: 10.1109/JSYST.2019.2919642

In a device-free human detection and tracking system using a received signal strength indicator (RSSI), the change in the RSSI pattern is monitored and analyzed to detect and track the human movement. The variation in measured RSSI signals is one of the major effects leading to significant detection and tracking error. To handle such a research problem, in this paper, we propose adaptive RSSI filtering methods designed by considering both the detection and tracking accuracy and the computational complexity. The novelty of our proposed filtering methods is that, to reduce the computational complexity, the measured RSSI input values are automatically filtered only when they have high variation levels; an appropriate threshold is set and used for the decision. Additionally, to increase the detection and tracking accuracy, the measured RSSI input values with different variation levels are filtered with different filtering levels adaptively. The proposed filtering methods are verified by the experiments, which have been carried out in an indoor environment. Various human movement patterns with different directions and speeds are tested. The experimental results show that, with our test scenarios, the proposed filtering methods can appropriately reduce the RSSI variation. They provide good detection and tracking accuracy, which is measured by the number of times the system can detect and track the human with the correct zone. The computational complexity measured by the number of mathematical operations, used by the proposed methods, is lower than comparative filtering methods.