Classification of passive sonar signals using a backpropagation neural network: simulation studies

PDF

Authors
  1. Cutmore, T.R.H.
  2. Arrabito, G.R.
Corporate Authors
Defence and Civil Inst of Environmental Medicine, Downsview ONT (CAN)
Abstract
For the past several decades passive sonar signals have been processed by the human operator using visual and/or auditory displays to achieve detection and classification of target vessels. Recently, a small number of studies have begun to apply neural network techniques to this domain of signal processing. Time series data presents challenges to these pattern recognition networks and preprocessing methods are often critical. In this paper, a passive sonar hydrophone simulation is outlined. Methods and results of applying the backpropagation algorithm with a 3-layer feedforward network to simulated passive sonar data are presented. In particular, different types of fourier preprocessing and network parameters are examined for their effects on convergence rate (learning speed) and classification performance. Finally, the extension of this work to real passive sonar data and potential pitfalls are discussed.
Report Number
DREA-TC-93-305-VOL-1-P-105 — CONTAINED IN 93-02664
Date of publication
01 Feb 1993
Number of Pages
23
DSTKIM No
93-02658
CANDIS No
131303
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: