Underwater acoustic targets classification using support vector machine

This paper introduces statistical learning and support vector machine (SVM) theory and a new SVM classification method, applied to the classification of underwater acoustical targets.

Abstract

Underwater target classification is a difficult task. Its performance depends mainly on the classifier designed using some exemplars. This paper introduces statistical learning and support vector machine (SVM) theory. Three SVM algorithms, linear SVM, non-linear SVM and multi-class SVM, are discussed. Based on it, a new SVM classification method is proposed. It is applied to the classification of underwater acoustical targets. Its classification performance is compared using same target data bank with traditional methods, such as k-nearest neighbor and neural network. Experiment results show that the proposed SVM classifier has better classification rate than traditional ones and advantages in selecting model, overcoming over-fitting and local minimum, etc.