舰船噪声听觉特征增强方法  

Listening feature enhancing algorithm for ship noise

作　　者：李训诰[1] 周利辉[1,2] 李馨子 

机构地区：[1]海军潜艇学院,山东青岛266041 [2]海军装备研究院,北京100171 [3]国家海洋局第一海洋研究所,山东青岛266041

出　　处：《声学技术》2017年第3期233-237,共5页Technical Acoustics

摘　　要：为改善新声呐兵听音训练效果,强化噪声样本中与目标属性紧密相关的听觉特征,将处理后的舰船噪声用于听音训练,提高舰船噪声样本利用率。提出的舰船噪声听觉特征增强模型及其实现方法主要包括四个步骤:噪声分解、子带分析、子带加权、噪声重建。首先采用多分辨分解将舰船噪声信号划分为若干子带,对子带进行逐个地听音分析和各种谱分析,选择特征信息稳定且丰富的子带进行强化,选取并微调各子带加权系数,采用多分辨分析理论重建舰船噪声,并根据重建噪声的功率谱、包络谱分析和听音分析结果调整加权系数,进而用于听音训练。仿真分析中,采用听音分析和谱分析比对原始噪声和重建噪声,验证了该方法的合理性和实用价值。In order to improve ship noise listening training for new sonar men, the listening features of the ship noise are enhanced, which are closely related to the properties of the targets. The processed signals are applied to listening training and the operation rate of ship noise samples is increased. The presented listening feature enhancement and its implementation method included 4 steps that are ship noise decomposition, sub-band analyzing, sub-band weighting and ship noise reconstruction. Firstly, multi-resolution is introduced to decompose ship noise into several sub-bands and every sub-band is analyzed through listening and various spectrum analyses. The sub-bands with stable and abundant information are selected and enhanced with power spectrum, DEMON （detection of envelope modulation on noise） spectrum and listening analysis. By fine turning the weighted coefficient of each sub-band, the ship noise is reconstructed with multi-resolution and applied to listening training. The comparison between original ship noise and reconstructed noise is done with listening analysis and spectrum analysis. It illustrates the rationality and application value.

关 键 词：舰船噪声 听觉特征增强 听音训练 分解重建 

分 类 号：TB556[理学—物理]