Clutter Reduction for ASW Using Automatic Aural Classification With a Coherent Source – FY2009 Report

PDF

Authors
  1. Hines, P.C.
  2. Murphy, S.
Corporate Authors
Defence R&D Canada - Atlantic, Dartmouth NS (CAN);Office of Naval Research, Arlington VA (US)
Abstract
This report presents the results from the first year of a three year research grant from the US Office of Naval Research (ONR). The project’s aim is to develop a robust classifier using aurally-based features that can discriminate active sonar target echoes from unwanted clutter echoes. A secondary objective is to examine the hierarchy (or more accurately, the interdependence) of classification and tracking to improve the integration of classification and tracking in future systems. During the first year of the project, an experiment was conducted on board NRV Alliance as part of the NURC Clutter09 sea trail. This enabled temporal robustness of the aural classifier to be examined by training the classifier using data collected during a 2007 field trial (Clutter07) and testing on data collected during Clutter09. The Fisher linear discriminant was used to rank the effectiveness of the classifier’s aural features. One of the most useful metrics to rate classifier performance is the area under the Receiver Operating Characteristic (ROC) curve, Az. The greater Az, the better the classifier, with a value of Az = 1 indicating ideal performance. The training ROC curve obtains a value of A z = 0.97 and the testing data obtains a value of A z = 0.90. These numbers are indicative of a very successful classifier. During the remaining two years of the research grant, tracker-classifier integration will be examined and the aural cl

Il y a un résumé en français ici.

Report Number
DRDC-ATLANTIC-ECR-2009-241 — External Client Report
Date of publication
01 Oct 2009
Number of Pages
30
DSTKIM No
CA033885
CANDIS No
533293
Format(s):
Electronic Document(PDF)

Permanent link

Document 1 of 1

Date modified: