A decision tree SVM classification method based on the construction of ship-radiated noise multidimension feature vector

A decision tree support vector machine (SVM) classification method based on the construction of ship-radiated noise multidimension feature vector is proposed in this paper. Aimed at three kinds of ship targets (class I submarine, class II warship and class III merchant ship) radiated noise, the subband distribution feature vectors of their 1½-spectrum and 2½-spectrum, and scale-energy feature vector of them based on wavelet transform are constructed respectively. And then a 55-dimension comprehensive feature vector of the ship-radiated noise is constructed. On this basis, a 24-dimension feature vector is obtained by using K-L transform for feature optimization. Finally, support vector machine technique is applied for the classification and it enhances the classification accuracy.