Lightweight Time Encoded Signal Processing for Vehicle Recognition in Sensor Networks

In this paper, we investigate a different approach to the vehicle classification task in wireless sensor networks. Instead of using traditional spectral or wavelet techniques to extract a feature vector, representative of each vehicle, we use a time-domain feature extraction method. The output of this coding procedure is a histogram-like matrix of fixed dimensions. These matrices can be used to train an artificial neural network (ANN) to classify different types of vehicles. The method is evaluated using data from a real world experiment, which contains acoustic and seismic recordings from two vehicles, a heavy wheeled truck (dragon wagon) and a tracked vehicle (assault amphibian vehicle). The dataset is available with the name Sitex02