针对基于最小能量统计的语音增强算法,用可靠的噪声段检测方法对噪声谱估计流程进行改进,提升噪声谱幅度估计的准确度。实验表明,此方法能有效地追踪背景噪声的变化情况,进而提升语音增强的效果。
An improved speech enhancement algorithm based on minimum statistics is proposed in this paper. The noise power estimation scheme is improved by using a reliable noise frame detection method. The experiment result shows that the proposed method can trace the variation of the noise floor efficiently, and lead to an improved speech enhancement result.