Recognition of marine acoustic target signals based on wave structure and Support Vector Machine

Abstract

According to research findings of speech acoustics, the timbre is applied to identify different types of targets. Since the information of timbre is indicated in the wave structure of time series, the feature of wave structure can be extracted to classify various marine acoustic targets. The method of feature extraction based on wave structure is studied. The nine-dimension feature vector is constructed on the basis of signal statistical characteristics, including zero-crossing wavelength, peek-to-peek amplitude, zero-crossing-wavelength difference, wave train areas and so on. And the Support Vector Machine (SVM) is applied as a classifier for two kinds of marine acoustic target signals. The kernel function is set Radial Basis Function (RBF). The penalty factor and parameter of RBF are properly selected by the method of combination of Differential Evolution (DE) and Particle Swarm Optimization (PSO), which helps to obtain better recognition rates than the grid search method.