No abstract

This paper describes a new criterion of speech recognition using an integrated confidence measure for minimization of the word error rate (WER). Conventional criteria for WER minimization obtain an expected WER of a sentence hypothesis merely by comparing it with other hypotheses in an n-best list. The proposed criterion estimates the expected WER by using an integrated confidence measure with word posterior probabilities for a given acoustic input. The integrated confidence measure, which is implemented as a classifier based on maximum entropy (ME) modeling, is used to get probabilities reflecting whether the word hypotheses are correct or incorrect. The classifier comprises a variety of confidence measures and can deal with a temporal sequence of them in order to attain a more reliable confidence. Our proposed criterion achieved a WER of 7.5% and a 2.6% improvement relative to conventional n-best rescoring methods in transcribing Japanese broadcast news under noisy field and spontaneous speech conditions.

SUMMARY The extraction of acoustic features for robust speech recognition is very important for improving its performance in realistic environments. The bi-spectrum based on the Fourier transformation of the third-order cumulants expresses the non-Gaussianity and the phase information of the speech signal, showing the dependency between frequency components. In this letter, we propose a method of extracting short-time bispectral acoustic features with averaging features in a single frame. Merged with the conventional Mel frequency cepstral coefficients (MFCC) based on the power spectrum by the principal component analysis (PCA), the proposed features gave a 6.9% relative lower a word error rate in Japanese broadcast news transcription experiments.

No abstract

No abstract