Syllable Detection in Continuous Speech,
Abstract
The report describes a syllable detection program which will provide sufficient knowledge about the presence of syllables in a speech passage to permit automatic stress contour analysis. The syllable detection program was designed for use in a top-down speech recognition system. Therefore, one of the significant features of the program is its ability to obtain an estimate on syllable placement without first attempting phonetic recognition. P-P amplitude and RMS intensity were separately used for the syllable detection work. 85.7% recognition rate was obtained from the use of P-P amplitude and 82.6% from RMS intensity. The results were discussed and error types analyzed. When both the amplitude and power spectrum measurements were used, the syllable detection rate can be improved up to 91.7%.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1974
- Accession Number
- ADA005718
Entities
People
- D. C. Sargent
- King Sun Fu
Organizations
- Purdue University