Energy efficient and robust CSIP algorithm in distributed wireless sensor networks

Abstract

This paper develops an energy efficient and robust collaborative signal and information processing (CSIP) algorithm and applies it to vehicle classification applications. The conventional algorithms collaboratively process all the time-series data from every node in the network. This signal–noise ratio (SNR)-based CSIP algorithm (SNRCSIP) collaboratively processes only the extracted features from part of the nodes. This algorithm efficiently reduces the energy consumption compared with the conventional algorithms by reducing the traffic. Apart from the energy efficiency, we demonstrate the robustness of the SNRCSIP algorithm by giving the high correct recognition ratio between the tracked vehicle and the wheeled vehicle with the acoustic features extracted by an improved form of mel filter bank (MFB), which is rarely applied in vehicle classification applications. Experimental results show that the SNRCSIP algorithm greatly reduces the energy consumption and achieves quite satisfied correct recognition ratio with the features extracted by the improved MFB.