基于能量和鉴别信息的语音端点检测算法

为提高实时通信中语音端点检测系统的性能,提出了一种基于能量和鉴别信息的端点检测算法。该算法利用帧信号的能量、子带信号的能量等参数,计算该帧信号与噪声帧基于子带能量分布概率的鉴别信息。算法通过利用鉴别信息,能够在包括语音帧在内的所有帧中更新噪声的能量,从而更准确地跟踪噪声能量的变化。实验结果表明:与基于能量的端点检测算法相比,该方法在信噪比变化比较剧烈的情况下仍然能够较准确地进行端点检测,在0~10 dB范围内变化的坦克噪声环境中,准确率比后者提高约24%。

Voice activity detection algorithm based on energy and discrimination entropy

A new algorithm based on the energy and discrimination information was developed to improve the performance of the voice activity detection system in real-time speech communications.The frame energy and the sub-band energy were used to calculate the discrimination information based on the sub-band energy distribution probabilities both for the current frame and the noise frame. The algorithm uses this discrimination information to update the noise energy in all frames including the speech frames,so it could better trace changes in the noise-energy.Tests show that the method gives a precise voice activity detection in the case of the seriously changed noise environment compared with the algorithm based on the energy.In noisy tank environments with noise level between 0 and 10 dB,the accuracy is increased by 24%.