摘要： 针对短时谱幅度最小均方误差估计(MMSE-STSA)语音增强算法在非平稳噪声环境下噪声抑制能力较差的问题，提出一种改进的MMSE-STSA算法。该算法在响度域进行谱幅度估计，并采取基于概率的语音判决方法。实验结果表明，相比传统的MMSE-STSA算法，改进算法在非平稳噪声环境下能够获得更好的语音质量。

关键词: 语音增强, 信噪比, 短时谱幅度最小均方误差估计, 响度域

Abstract: Due to the poor performance of traditional Minimum Mean Square Error Short-Time Spectral Amplitude estimator(MMSE-STSA) algorithm in non-stationary noise environment, this paper proposes an improved MMSE-STSA speech enhancement algorithm, which is based on a loudness-domain MMSE estimator in the frequency domain, and uses a probability-based voice decision method. Simulation result indicates that it has better performance than traditional MMSE-STSA algorithm in non-stationary noise environment.

Key words: speech enhancement, SNR, Minimum Mean Square Error Short-Time Spectral Amplitude estimator(MMSE-STSA), loudness domain

中图分类号: 

• TN911.72