Selection and parameter optimization of SVM kernel function for underwater target classification

The identification and classification of noise sources in the ocean has become a key task of modern underwater acoustic signal processing and because of the ever changing and complicated oceanic environment, underwater target classification has become a demanding task. An underwater acoustic target classification system identifies the acoustic target from the characteristic acoustic signature. The characteristic acoustic signatures are patterned by feature recognition algorithms operating on data captured by hydrophone. In this paper, an SVM classifier works on the acoustic signatures of four different target types. The performance of the classifier depends on a variety of factors, of which SVM parameter tuning is very important. Several attempts have been made for automatic kernel selection and parameter optimization, including meta-heuristic algorithms such as genetic algorithms (GA) and particle swarm optimization (PSO). This paper attempts towards selection of SVM parameters, kernel and kernel parameter optimization using the BAT algorithm. The results indicate higher classification accuracy when compared to PSO based selection and optimization.