提出了一种适用于复杂加性噪声环境的语音端点检测方法。通过对噪声类型的归纳,建立自适应的平稳噪声模型,并根据模型搜索信号能量非平稳的区域。然后基于浊音在频域上的谐波结构,在此区域内检测浊音,从而排除非平稳噪声的干扰。最后根据信号能量精确搜索语音起止点。与目前典型端点检测算法的对比实验表明,在大多数复杂噪声环境下,该算法具有较好的准确率。 A speech endpoint detection method suitable for complex additive noise environment is proposed. By inducting the noise types, an adaptive stationary noise model is established and the non-stationary region of the signal energy is searched according to the model. Then based on the harmonic structure of the voiced sound in the frequency domain, the voiced sound is detected in this area to eliminate the interference of the non-stationary noise. Finally, according to the signal energy to accurately search for voice start and stop points. Compared with the current typical endpoint detection algorithm, it shows that this algorithm has better accuracy in most complex noise environments.