Multi-target classification and tracking using wireless sensor networks

Abstract

This thesis presents an acoustic classification and multi-sensor tracking algorithm with WSNs for multiple targets is suggested. The goal for this study is to classify and track the traces of moving multi-targets with their acoustic characteristics and received signal strength indicator. The thesis includes unique method to select features from the raw acoustic signal which contains both time and frequency domain analysis. For localization, Gaussian process based algorithm for estimating unidentified traces from received signal strength indicator is presented. In addition, the unique method for labeling unidentified traces with appropriate target is introduced. By using suggested algorithm, the classifier shows more accurate and faster responding performance than the classifiers which use only frequency domain input. Experimental results with show the satisfactory performance of the proposed algorithm.