REJECTION BASED ON A POSTERIORI PROBABILITY ESTIMATED BY MLP WITH APPLICATION FOR MANDARIN VOICE DIALER ON ASIC
Abstract:
High performance mandarin voice dialer is much more difficult than its English counterpart to achieve, especially on inexpensive hardware as ASIC. One way to improve its performance is to incorporate rejecters into the system. In our study, an MLP based postprocessor, an a posteriori probability estimator, is applied after HMM Viterbi recognizing. Poor utterances, which are recognized by HMMs but have low a posteriori probability, are then rejected. Rejecting 4.9 % of all the testing utterances, the MLP rejecter boost the HMM-based system's single digit accuracy from 97.1 % to 99.6 % for the Mandarin voice dialer, a ten-syllable speaker independent task. The performance is better than those of rejection based on linear discrimination, anti-digit models or likelihood ratio. 1. INTRODCUTION
Citations
| 193 | Neural network classifiers estimate Bayesian aposteriori probabilities – Richard, Lippmann - 1991 |
| 34 | Automatic recognition of keywords in unconstrained speech using hidden markov models – Wilpon, Rabiner, et al. - 1990 |
| 21 | High-Performance Alphabet Recognition – Loizou, Spanias - 1996 |
| 18 | Discriminative utterance verification for connected digits recognition – Rahim, Lee, et al. - 1997 |
| 4 | A two pass classifier for utterance rejection in keyword spotting – Sukkar - 1993 |
| 4 | Rejection of extraneous input in speech recognition applications, using multi-layer perceptrons and the trace of HMMs – Mathan, Miclet - 1991 |
| 2 | Rejection for connected digit recognition based on GPD segmental discrimination – Sukkar - 1994 |
| 1 | Rejection techniques for Digit Recognition – Villarrubia, Acero |
| 1 | Mandarin Digit Speech Recognition: State of the Art – Gu, Liu |
