基于样本熵的语音信号检测算法

船舶自身发出的巨大噪声,严重影响了通讯设备通讯端船舶工作者的正常语音交流和通讯,须要进行语音信号增强,将语音信号从背景噪声中检测分离出来。为此,文章在分析语音信号与船舶背景噪声信号时域序列复杂性基础上,提出基于信号序列样本熵的语音信号检测算法。该算法将含噪信号进行分帧后,首先计算每帧信号的样本熵,并根据自适应预设阈值对信号进行初始判断；然后,利用信号通常持续一定时间的连续性,对信号初始判断进行平滑,得到信号的最终判别类别,实现对语音信号的检测。经过大量实测实验数据的验证,环境噪声信号的辨别准确度为 90.33%,语音信号为 95.05%,该算法取得了较好的判别精度。

Detection Algorithm of Voice Signals Based on the Sample Entropy

The ship environment noise from its relevant equipment influences the communication. Based on the analysis ofthe time-domain sequence complexity of the voice signal and environment noise, the paper proposes a new detection algo?rithm of voice signals based on the sample entropy. The sample entropy of each signal frame is calculated and the signal isinitially judged according to the adaptive threshold. Based on the continuity of the signal that always lasts for some time,the final results are obtained after smoothing the initial judgment of the signal. The algorithm is examined on a large mea?sured data, with correct discrimination accuracy of 90.33% for ship-radiated noise and that of 95.05% for voice. Experi?mental results show that the efficiency of the algorithm is exceptionally good.