极区潜标锚系噪声检测抑制及噪声谱级年度统计分析
Self-noise detection and reduction in polar submerged moorings and annual statistical analysis of noise spectrum level
-
摘要: 极区海域声学环境以包含大量高强度瞬态冰噪声事件为显著特征, 被动声学监测系统中的锚系噪声与冰源噪声的混淆问题影响数据可靠性, 为此提出了一种针对锚系噪声的协同检测与抑制方法。首先建立基于K均值聚类与主成分分析的检测模型, 通过谱级特征提取实现噪声检测; 进而开发改进中值滤波的谐波分离算法, 结合自适应分离因子与时频域软掩膜动态重建机制进行干扰抑制。实验结果表明, 该方法能有效抑制典型锚系噪声频段平均谱级约10 dB, 并保留环境声源。研究发现锚系噪声产生与设备温度、深度的周期变化一致, 说明锚系系统抖动是产生锚系噪声的主要原因。所提方法通过建立锚系噪声检测抑制模型, 解决了极区声学数据中特定锚系噪声与自然声源的解耦难题, 其检测精确率达88.9%, 误检率低于1.1%, 为极区环境声学监测提供了可靠的技术手段和理论依据。Abstract: The acoustic environment in polar sea areas is characterized by numerous high-intensity transient ice noise events, and the confusion between mooring noise and ice-generated noise in passive acoustic monitoring systems undermines data reliability. This study proposes a collaborative detection and suppression method for mooring noise. First, a detection model based on K-means clustering and principal component analysis is developed to extract spectral-level features for noise detection. Then, an improved median filtering-based harmonic separation algorithm is introduced, combining an adaptive separation factor with a time-frequency domain soft mask dynamic reconstruction mechanism for interference suppression. Experimental results demonstrate that the method effectively reduces the average spectral level by approximately 10 dB in typical mooring noise frequency bands while preserving environmental sound sources. The study also finds that the occurrence of mooring noise exhibits periodic variations consistent with equipment temperature and depth, and supplementary pool experiments confirm that the mechanical movement of the equipment’s vibration damping apparatus is the primary cause of the noise. Overall, this method addresses the decoupling challenge between specific mooring noise and natural sound sources in polar acoustic data, achieving a detection accuracy of 88.9% with a false alarm rate below 1.1%, thereby providing a reliable technical means and theoretical basis for polar environmental acoustic monitoring.