PDF

Click for pdf file

Title

Integration of neural classifiers for passive sonar signals

Authors

Summary

The identi cation and classi cation of underwater acoustic signals is an extremely di- cult problem b ecause of low SNRs and a high degree of variability in the signals emanated from the same typ e of sound source.Sincedi erent classi cation techniques have di erent inductive biases, a single metho d cannot give the b est results for all signal typ es.Rather, more accurate and robust classi cation can obtained by combining the outputs evidences of multiple classi ers based on neural network and or statistical pattern recognition tech- niques.In this pap er, ve approaches are compared for integrating the decisions made by networks using sigmoidal activation functions exhibiting global resp onses with those made by lo calized basis function networks.Thesemetho ds are compared using realistic o ceanic data.The

Keywords

underwater acoustic, variability, sourcesince, techniques inductive, accurate, evidences multiple, statistical, recognition, approaches decisions, decisions networks, networks sigmoidal, sigmoidal activation, activation functions, function networksthese, realistic, entropy-based weighting, weighting individual, combination, mycinthe,

Links

001 002