Variance Compensation for Recognition of Reverberant Speech with Dereverberation Preprocessing

Marc Delcroix, Shinji Watanabe, Tomohiro NakataniRobust Speech Recognition of Uncertain or Missing Data 2011 · OpenReview

Abstract

The performance of automatic speech recognition is severely degraded in the presence of noise or reverberation. Much research has been undertaken on noise robustness. In contrast, the problem of the recognition of reverberant speech has received far less attention and remains very challenging. In this chapter, we use a dereverberation method to reduce reverberation prior to recognition. Such a preprocessor may remove most reverberation effects. However, it often introduces distortions, causing a static and dynamic mismatch between speech features and the acoustic model used for recognition. Acoustic model adaptation could be used to reduce this mismatch. However, conventional model adaptation techniques assume a static mismatch and may therefore not cope well with a dynamic mismatch arising from dereverberation. This chapter introduces a novel model adaptation scheme that is capable of managing both static and dynamic mismatches. We introduce a parametric representation of Gaussian variances of the acoustic model that includes static and dynamic components. Adaptive training is used to optimize the variances in order to realize an appropriate interconnection between dereverberation and a speech recognizer.
BibTeX
@inproceedings{delcroix2011variance,
  author = {Marc Delcroix and Shinji Watanabe and Tomohiro Nakatani},
  title = {{Variance Compensation for Recognition of Reverberant Speech with Dereverberation Preprocessing}},
  booktitle = {Robust Speech Recognition of Uncertain or Missing Data},
  year = {2011},
  url = {https://openreview.net/forum?id=wOCrhYuldw},
}

Loading related content...

Figure