The temporal and spectral features of the acoustic signatures of aircraft and highway vehicles are used to train two types of neural networks. The detection and classification performances of the networks are then evaluated using an independent set of data. The types of networks evaluated are (i) feedforward with backpropagation training, and (ii) probabilistic nets with maximum-likelihood training. The results demonstrate accurate classification as to type of vehicle during aircraft takeoff or under load conditions for heavy trucks, depending on the signal-to-noise ratio. Neural nets appear to offer a promising vehicle classifier for monitoring airport and highway traffic for compliance with noise regulations.
Feedforward networks employing the backward propagating delta rule for error correction were tested utilizing simulated target signatures and noise to provide insight into the network learning process. Network training histories and weight evolutions were studied for alternating signal and noise input vectors for two network architectures. Contour plots of the input-to-hidden layer weights clearly indicate the relationship between the evolving features of the network weights as they respond to the input signatures during the learning process. Singular value decomposition of the input-to-hidden layer transfer matrix provides insight into the similarities of a trained neural network and a classical matched filter.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.