Adaptive speech rate conversion without extension of input data duration, using speech interval detection
US6236970A · kind A · utility
8Cited by
0References
8Claims
0Family size
Assignee
Inventors
Key dates
| Filing date | Dec 22, 1998 |
| Grant date | May 22, 2001 |
| Priority date | — |
| Expiry date | Dec 22, 2018 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG10L2025/786
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A speech-rate converter slowing down input speech regularly monitors the data length of the input speech and the previously estimated extended output data length for the current rate scaling factor, computing new output data length estimates. The conversion rate is adaptively modified depending on the time lag between input and output speech so as to make input and output data lengths consistent without skipping any spoken input portions. Input signal power is monitored to discriminate speech and non-speech intervals, and the portions of input non-speech intervals exceeding a conversion-rate-dependent duration are deleted.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.