| | |
Summary: COMBINING FEATURE COMPENSATION AND WEIGHTED VITERBI DECODING FOR
NOISE ROBUST SPEECH RECOGNITION WITH LIMITED ADAPTATION DATA
Xiaodong Cui and Abeer Alwan
Department of Electrical Engineering
University of California, Los Angeles, CA 90095
Email: xdcui@icsl.ucla.edu, alwan@icsl.ucla.edu
ABSTRACT
Acoutic models trained with clean speech signals suffer in the
presence of background noise. In some situations, only a limited
amount of noisy data of the new environment is available based
on which the clean models could be adapted. A feature compen-
sation approach employing polynomial regression of the signal-to-
noise ratio (SNR) is proposed in this paper. While clean acoustic
models remain unchanged, a bias which is a polynomial function
of utterance SNR is estimated and removed from the noisy fea-
ture. Depending on the amount of noisy data available, the algo-
rithm could be flexibly carried out at different levels of granular-
ity. Based on the Euclidean distance, the similarity between the
residual distribution and the clean models are estimated and used
as the confidence factor in a back-end Weighted Viterbi Decoding
|
