Abstract: Based on Bark wavelet transform, a voice activity detection algorithm is proposed to improve the performance in noisy environment. This paper introduces a scale parameter into the nonlinear mapping, utilizes the singularity of the Bark wavelet transform coefficients in different sub-bands to construct a VAD parameter, and effective threshold selection rules are validated by simulations. Experiment results show that the accuracy of this algorithm is evidently heightened, and its robustness is enhanced, compared to some traditional algorithms such as short-time energy and cepstral distance.

Key words: voice activity detection, Bark wavelet transform, expansion scale parameter, parameter extraction

摘要： 为提高噪声环境下端点检测算法的性能，提出一种基于Bark小波变换的语音端点检测算法。在Bark小波变换的非线性映射中引入伸缩尺度参数，提取不同子带上信号的小波变换系数，利用Bark小波的奇异性构造端点检测参数，并通过仿真验证得到有效的阈值选择规则。实验结果表明，与传统的短时能量法和倒谱距离测度法相比，该算法的检测准确率明显提高，具有较强的噪声鲁棒性。

关键词: 语音端点检测, Bark小波变换, 伸缩尺度参数, 参数提取

CLC Number: 

• TN912.34