NEURAL NETWORKS FOR INDEPENDENT RANGE AND DEPTH DISCRIMINATION IN PASSIVE ACOUSTIC LOCALIZATION

PDF

Authors
  1. Zakarauskas, P.
  2. Ozard, J.M.
  3. Brouwer, P.
Corporate Authors
Defence Research Establishment Pacific, Victoria BC (CAN);Datavision Computing Services Ltd, Victoria BC (CAN)
Abstract
Two feedforward neural networks with one hidden layer each were trained using a fast backpropagation algorithm to determine the position of an acoustic source in a waveguide. One network was trained to localize the source in depth while the other was trained independently to localize in range. The output layer consisted of one unit for each possible range or depth of the source. The networks were trained with a signal-to-noise ratio (S/N) of 50 dB and tested with patterns generated with S/N ranging from O dB to 20 dB. The performance of the neural networks (NNs) was compared with that of a nearest-neighbour classifier. Evaluation of the processors was done in the context of an estimation problem, i.e. by measuring the root-mean-squared (rms) error of the processors' estimates. The NNs turned out to be less resistant to noise than the conventional processor, but were faster. An explanation is given as to why multilayered feedforward neural networks cannot in general achieve the performances of optimum classifiers.
Report Number
DREA-TC-93-305-VOL-1-P-51 — @Conference Paper; CONTAINED IN 93-02664
Date of publication
01 Feb 1993
Number of Pages
17 (p51-67)
DSTKIM No
93-02661
CANDIS No
131300
Format(s):
Microfiche filmed at DSIS;Originator's fiche received by DSIS;Document Image stored on Optical Disk

Permanent link

Document 1 of 1

Date modified: