A minimum-mean-square-error noise reduction algorithm on Mel-frequency cepstra for robust speech recognition
A minimum-mean-square-error noise reduction algorithm on Mel-frequency cepstra for robust speech recognition
复制标题
DOI：
10.1109/icassp.2008.4518541
复制
发表时间：
2008-05
期刊：
影响因子：
--
通讯作者：
Dong Yu;L. Deng;J. Droppo;Jian Wu;Y. Gong;A. Acero
中科院分区：
文献类型：
--
作者：
Dong Yu;L. Deng;J. Droppo;Jian Wu;Y. Gong;A. Acero
提出了一种基于最小均方误差(MMSE)准则的MFCC(MFCC)非线性特征域降噪算法，用于环境鲁棒性语音识别。与Ephraim和Malah(E&M)(1985)提出的对数谱幅度的MMSE增强不同，本文提出的新算法发展了适用于滤波器组输出功率谱幅度的抑制规则，并直接应用于MFCC，使其在抗噪语音识别中明显更有效。新算法中的噪声方差包含了E&M算法中缺失的一个重要项，该项是由干净语音和混合噪声之间的瞬时相位不同步引起的。在标准Aurora-3任务上的语音识别实验表明，相对于ICSLP02基线，错误率降低了48%，相对于倒谱平均归一化基线，错误率降低了26%，相对于传统的E&M LOG-MMSE噪声抑制器，错误率降低了13%。新算法也比E&M噪声抑制器高效得多，因为MEL频率滤波器组中的通道数(在我们的情况下是23个)比FFT域中的区段数(256)要少得多。实验结果还表明，在完全匹配和中等不匹配的设置下，我们的算法的性能略优于ETSI AFE。
We present a non-linear feature-domain noise reduction algorithm based on the minimum mean square error (MMSE) criterion on Mel-frequency cepstra (MFCC) for environment-robust speech recognition. Distinguishing from the MMSE enhancement in log spectral amplitude proposed by Ephraim and Malah (E&M) (1985), the new algorithm presented in this paper develops the suppression rule that applies to power spectral magnitude of the filter-banks' outputs and to MFCC directly, making it demonstrably more effective in noise-robust speech recognition. The noise variance in the new algorithm contains a significant term resulting from instantaneous phase asynchrony between clean speech and mixing noise, missing in the E&M algorithm. Speech recognition experiments on the standard Aurora-3 task demonstrate a reduction of word error rate by 48% against the ICSLP02 baseline, by 26% against the cepstral mean normalization baseline, and by 13% against the conventional E&M log-MMSE noise suppressor. The new algorithm is also much more efficient than E&M noise suppressor since the number of the channels in the Mel-frequency filter bank is much smaller (23 in our case) than the number of bins in the FFT domain (256). The results also show that our algorithm performs slightly better than the ETSI AFE on the well-matched and mid-mismatched settings.