Bispectrum based feature extraction and classification of radiation noises from underwater targets
Peng Yuan · Harbin Gongcheng Daxue Xuebao/Journal of Harbin Engineering University · 2003
For the radiation noises from underwater targets apparently contain non-Gaussian ingredients, their non-Gaussion features were studied through high-order cumulants. 65 dimensional bispectrum features were extracted from different targets through bispectrum estimation and WALSH dimensionality reduction, which indicates that the features greatly help the classification of noise signals from underwater targets, and efficiently restrains the coloured Gaussian noises. 92% of radiation noise signals from 6 types of underwater targets were properly identified.