Feature enhancement of reverberant speech by distribution matching and non-negative matrix factorization

@article{Keronen2015FeatureEO,
  title={Feature enhancement of reverberant speech by distribution matching and non-negative matrix factorization},
  author={Sami Keronen and Heikki Kallasjoki and Kalle J. Palom{\"a}ki and Guy J. Brown and Jort F. Gemmeke},
  journal={EURASIP J. Adv. Sig. Proc.},
  year={2015},
  volume={2015},
  pages={76}
}
This paper describes a novel two-stage dereverberation feature enhancement method for noise-robust automatic speech recognition. In the first stage, an estimate of the dereverberated speech is generated by matching the distribution of the observed reverberant speech to that of clean speech, in a decorrelated transformation domain that has a long temporal context in order to address the effects of reverberation. The second stage uses this dereverberated signal as an initial estimate within a non… CONTINUE READING