摘要
为了在海洋环境中检测有效被动鱼声信号段,实现鱼声信号的识别,采用稀疏分解算法提取相干比特征值实现鱼声端点检测技术。该算法在训练阶段提取不同信噪比条件下干净的被动鱼声、海浪噪声的特征值作为测试声目标特征,在检测阶段提取移动含噪信号段与测试特征做欧式距离识别分类,将识别对象分成两类,最后采用门限判决方法实现端点的检测。实验结果表明,相比较于功率谱特征提取算法,该算法在小信噪比条件下可以准确实现有效信号段的检测。
To detect and recognize passive fish acoustic signals from noisy marine environment, sparse decomposition is used to realize the endpoint detection by coherent ratio feature. This algorithm extracts feature of fish and wave acoustic as test object under different signal-to-noise （SNR） ratios in the training; feature from noisy signal segment is extracted in the testing to classify test object; finally, it is realized by threshold endpoint detection method. Experimental results show that the algorithm in low SNR can accurately realize effective signal segments, compared with the power spectrum feature extraction algorithm.
出处
《热带海洋学报》
CAS
CSCD
北大核心
2015年第4期48-53,共6页
Journal of Tropical Oceanography
基金
国家自然科学基金(51109218)
江苏省自然科学基金(BK20130245)
江苏省产学研联合创新资金研究项目(BY2014040)
常州工学院自然科学基金(YN1311)
关键词
被动鱼声
端点检测
稀疏分解
小信噪比
passive fish acoustic
endpoint detection
sparse decomposition
low signal-to-noise ratio
