Passive Sonar Signals Classification using Fusion of Short-Time Fourier Transform and timbre Features

Detection and classification of Marine vessels based on their acoustic radiated noise is an important part of sonar systems. I this paper passive sonar targets classification algorithms is reviewed and a new algorithm is proposed. LDA feature extraction algorithm and Fusion of short time Fourier and timbre features is used in proposed algorithm which is called STFTLDA-Timb. Extracted features of proposed algorithm, are highly discriminant and classification accuracy of proposed algorithm is 8.45% better than STFT based classification algorithm. Obtained results of real data passive sonar classification show that classification accuracy of proposed algorithm is better than some common classification algorithms like statistical classifiers, neural networks and Ensemble Learning algorithms.