Speech Signal Endpoint Detection Using Hidden Markov Models
A major cause of errors in automatic speech recognition system is the inaccurate detection of the beginning and ending boundaries of test and reference patterns. Separation of speech and silence segments in automatic speech recognition algorithms occupies a fundamental position. The improper demarca...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Journal Quality and Technology Management
2006
|
| Online Access: | http://eprints.utm.my/id/eprint/8757/1/JQTM-v2-n1.pdf http://eprints.utm.my/id/eprint/8757/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Teknologi Malaysia |
| Language: | English |
| Summary: | A major cause of errors in automatic speech recognition system is the inaccurate detection of the beginning and ending boundaries of test and reference patterns. Separation of speech and silence segments in automatic speech recognition algorithms occupies a fundamental position. The improper demarcation of these segments reduces system’s efficiency, since;the system has to execute processing on the portion of segments which are not needed. A comprehensive evaluation of ASR systems showed that more than half of the recognition errors are caused due to wrong word boundary detection [1][2] Therefore the desired
characteristics for an endpoint detection are reliability, robustness, accuracy, adaptation,simplicity and real time processing etc[3]. This paper discusses a robust algorithm for the detection of speech and silence segments in an input speech signal based on Hidden Markov Models |
|---|