摘要
针对宽吻海豚Click信号检测提出了一种在信号时频图中基于Gabor滤波器的检测方法。该方法首先对声信号进行分段处理,计算每一段信号的时频图;然后设计Gabor滤波器,提取时频图中垂直方向的线条;对Gabor滤波处理后的时频图进行自适应阈值处理,提取时频图中能量较强的区域;最后通过连通域分析确定Click信号的位置.仿真合成不同信噪比的测试信号,本文算法在Click信号和背景噪声平均功率比为15 dB的情况下,Click信号的找全率达到了99%,错误率为0%;对实际采集的声信号进行Click信号检测,找全率为100%。本文方法预期为海豚观测和海豚生物学行为的研究提供一定的技术支持。
A method based on Cabor filter is proposed for the detection of Pulses of Bottlenose dolphin （Tursiops truncates）. First, the acoustic signal is divided into frames, and the spectrogram of each frame is calculated. Then Gabor filter is designed to extract the vertical lines in the spectrogram. Adaptive threshold is used to detect the regions where pulses may be located in the spectrogram. Finally, the positions of the pulses are determined by the connected component analysis. Simulating test signal of different signal to noise ratio （SNR）, the detection rate is 99% and the error rate is 0% when SNR is 15 dB. Detecting pulses in real world acoustic signal, the detection rate is 100%. The method is expected to provide technical support for dolphin observation and the study of dolphins＇ biological behavior.
出处
《声学学报》
EI
CSCD
北大核心
2017年第4期445-450,共6页
Acta Acustica
