SVM based Classification for Underwater Transient Signals in Ocean Background Noise

摘要

In this paper, new method for classification of underwater transient mechanical signals was proposed. The classifier uses SVM algorithm and 16 orders LPC coefficients as feature vectors. The proposed classifier is composed of two steps. The mechanical signals are separated from biological signals at the first SVM classifier and then classification of mechanical signal is preformed at the second SVM. For experiment, three kinds of underwater biological signals and two kinds of mechanical signals were used. The recognition rate was higher than 90％ at high Signal to Noise Ratio (SNR) when clean signal used. When underwater ambient noise was added, the recognition rate dropped but was better than the Bayesian classifier.