Discriminative uncertainty estimation for noise robust ASR

Dung Tien Tran (1) , Emmanuel Vincent (2) , Denis Jouvet (2)

1 PAROLE - Analysis, perception and recognition of speech
2 MULTISPEECH - Speech Modeling for Facilitating Oral-Based Communication

We consider the problem of uncertainty estimation for noise-robust ASR. Existing uncertainty estimation techniques improve ASR accuracy but they still exhibit a gap compared to the use of oracle uncertainty. This comes partly from the highly non-linear feature transformation and from ad- ditional assumptions such as Gaussian distribution and independence between frequency bins in the spectral domain. In this paper, we propose a method to rescale the estimated feature-domain full uncertainty covariance matrix in a state-dependent fashion according to a discriminative criterion. The state-dependent and feature index-dependent scaling factors are learned from development data. Experimental evaluation on Track 1 of the 2nd CHiME challenge data shows that discriminative rescaling leads to better results than generative rescaling. Moreover, discriminative rescaling of the Wiener uncertainty estimator leads to 12% relative word error rate reduction compared to discriminative rescaling of the alternative estimator in [1]

Mots clés: IndexTerms—Automaticspeechrecognition; discriminativeadaptation; uncertaintyhandling; noisero-bustness

Domaines: Son [cs.SD]

Dung Tien Tran, Emmanuel Vincent, Denis Jouvet. Discriminative uncertainty estimation for noise robust ASR. 40th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2015, Apr 2015, Brisbane, Queensland, Australia. ⟨hal-01103969⟩