Abstract
To improve the effect of sentence-evel English audio segmentation, a method of sentence boundary detection based on dual-threshold is proposed. With consideration of the characteristics of normative waveform, small noise and stable speed for English audio such as VOA and BBC, the method in this paper detects the sentence boundary of English audio via its quiet energy threshold and quiet delay threshold, and the corresponding audio text is used to revise the segmentation result. Experiments on special English of VOA show that the recall rate of segmentation exceeds 96% and the precision rate exceeds 94% by using the dual-threshold method only. After revision via the corresponding text, the precision rate can be improved further.
| Original language | English |
|---|---|
| Pages (from-to) | 259-263 |
| Number of pages | 5 |
| Journal | Harbin Gongye Daxue Xuebao/Journal of Harbin Institute of Technology |
| Volume | 42 |
| Issue number | 2 |
| State | Published - Feb 2010 |
| Externally published | Yes |
Keywords
- Audio segmentation
- Boundary detection
- Dual-threshold
Fingerprint
Dive into the research topics of 'Boundary detection of special English audio sentence based on dual-threshold'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver