Ship noise demodulation line spectrum fusion feature extraction based on the wavelet packet

Feature extraction of ship radiated-noise is key technique of underwater acoustic target recognition. Firstly, the mul-subspace speciality of ship noise is studied by using the wavelet packet analysis. And multi-subspace demodulation spectrum of ship noise is obtained. Secondly, the evaluation method of multi-subspace demodulation spectrum confusion degree is studied by using information entropy. Thirdly, the extraction method of ship noise demodulation line spectrum fusion feature is put forward by using the wavelet packet reconstruction algorithm and information entropy evaluation criterion. The comparative result shows that the demodulation line spectrum fusion feature based on the wavelet packet has higher signal-to-noise ratio and better recognition ability than the demodulation line spectrum feature of primary signal. And the demodulation line spectrum fusion feature based on the wavelet packet could improve underwater acoustic target recognition rate effectively.