APPLICATION OF NEURAL COMPUTING TO SONAR CONTACT DATA MANAGEMENT

PDF

Authors
  1. Brobeck, C.
  2. Pilkington, C.
  3. Cunningham, J.
Corporate Authors
Defence Research Establishment Atlantic, Dartmouth NS (CAN);Array Systems Computing Inc, Downsview ONT (CAN)
Abstract
In this report we put forward a method to assist in the automated detection and maintenance of passive sonar tracks. The technique employed is based on a neural network strategy known as Growing Cell Structures. This strategy is closely linked to Kohonen's feature map techniques and Radial Basis Function networks. Among other advantages, it demonstrates fast learning, an ability to distribute itself evenly over complex topologies, and the ability to deal with sharp boundaries and discrete regions. In the report we develop a design for a system to initialize track following mechanisms. The work is then extended to following and loss-of-lock scenarios. The effectiveness of the design and implementation is demonstrated with simulated data at several ratios of signal-to-noise. The report also contains a discussion of why this method was selected over the alternative techniques considered in the first phase of this project. The interim report concerning the first phase is included as an appendix. The report concludes with a discussion of the possible extensions and improvements to the current work. A summary of the software is also included as an appendix.
Keywords
ANN (Artificial Neural Networks);Growing Cell Structures;Radial Basis Function;Loss Of Lock
Report Number
DREA-CR-93-479 — Contractor Report
Date of publication
01 Oct 1993
Number of Pages
71
DSTKIM No
96-02285
CANDIS No
469953
Format(s):
Document Image stored on Optical Disk;Hardcopy

Permanent link

Document 1 of 1

Date modified: