摘要
本文主要研究语音增强系统在短时频域内,以对数最小均方误差作为语音幅值估计的准则.对含噪语音进行消噪处理。首先推导了公式,并从理论上作了分析研究.在此基础上作了大量实验,与类似的最小均方误差短时谱幅值估计法(MMSE STSA)作了性能比较.实验结果表明,由MMSE LOG-STSA增强方法得到的语音输出的质量优于由MMSA STSA方法得到的质量.语音的失真度和清晰度均有显著改善,同时大大降低了语音中的“音乐噪声”.
This paper focuses on the speech inhancement system which utilizes a minimum mean-square error log short-time spectral amplitude estimator. The system can perform the noise-cancel processing of noisy speech. Firstly we derive the Minimum Mean-Square Error LOG Short-Time Spectral Amplitude (MMSE-LOG-STSA) estimator. And then we analyse the performance of the proposed estimator. Based on the algorithm, we have done widely the experiments using this method, and compared this method to the Minimum Mean-Squate Error Short-Time Spectral Amplitude (MMSE-STSA) method. The experimental results show that the Output speech using the MMSE LOG-STSA system is prior to that of MMSE STSA method in its quality. The distortion and clarity of speech are significantly improved, and the 'music noise' in speech is dramatically decreased.
出处
《南京邮电学院学报》
北大核心
1991年第3期1-6,共6页
Journal of Nanjing University of Posts and Telecommunications(Natural Science)
关键词
语音信号处理
语音增强
幅值估计
Speech processing
Noise control
Frequency domain analysis
