Abstract
In this paper, a two-stage recognition scheme, phonetic recognition followed by prosodic recognition, is established. In the phonetic recognition process, 21 INITIAL and 37 FINAL context-independent HMM's are used to construct the phonetic recognizer. In the prosodic recognizer, 175 context-dependent prosodic HMM's are used to model the complicated tone behavior for all possible tone concatenations. Five anti-prosodic HMM's, each corresponding to one lexical tone, are constructed to enhance the discriminability among prosodic HMM's. This system was evaluated in a speaker-dependent mode on a vocabulary size of thirty thousand words. The experimental results show that the recognition rate was improved from 80.3% to 86.7% using the prosodic information.
| Original language | English |
|---|---|
| Pages | 253-256 |
| Number of pages | 4 |
| Publication status | Published - 1997 |
| Event | Proceedings of the 1997 IEEE TENCON Conference. Part 1 (of 2) - Brisbane, Australia Duration: 1997 Dec 2 → 1997 Dec 4 |
Other
| Other | Proceedings of the 1997 IEEE TENCON Conference. Part 1 (of 2) |
|---|---|
| City | Brisbane, Australia |
| Period | 97-12-02 → 97-12-04 |
All Science Journal Classification (ASJC) codes
- Computer Science Applications
- Electrical and Electronic Engineering