船舶辐射噪声盲分离技术仿真研究

基于盲分离的思想，船舶辐射噪声可通过盲源分离技术重构出相应的原始辐射信号。仿真实验研究了两种情况的源分离：从单船舶辐射噪声信号谐波成分的分离；多船舶辐射混合信号中分离出每条船只对应的原始辐射信号。采用理论模型产生的模拟数据与实际录音数据进行的实验，得到了满意的结果。文中采用神经网络自适应在线学习算法，算法稳定性好、收敛速度快，在极低的信噪比下仍具有很好的分离效果。

Simulation Study on Blind Separation Techniques for Ship Radiated Noises

Based on the idea of blind source separation, this paper shows that the original source signals of interest could be reconstructed from the hybrid complex underwater signals by a blind source separation technique. Two simulation experiments have been carried out in the paper: (1) For one ship, the components corresponding to different harmonic vibration sources in a ship could be decomposed from its radiated signals. (2) For multi-ship existence, each of the ship-radiated noises in the mixture signals could be separated by the blind source separation technique. Based on two kinds of simulated data produced by theoretic models and real recorded data, the simulation experiments achieved satisfactory results. It was shown that the adaptive on-line neural learning algorithm used in the paper is stable, fast and robust to noise.