14th International Congress of Phonetic Sciences (ICPhS-14)
San Francisco, CA, USA
A new acoustic-phonetic feature-and knowledge-based approach for the detection and classification of stop consonants in speakerindependent continuous speech is proposed. A system is built which automatically extracts stop consonants from continuous speech. The detected stop consonants are then passed to the classification system, which classifies them according to their voicing and place of articulation. The system combines multiple static and dynamic features in a multi-level decision process. It uses an auditory-based front-end processing and incorporates new algorithms for the extraction and manipulation of the acoustic-phonetic features that proved to be useful in the recognition process. The system is tested on 300 sentences from 30 different speakers of the TIMIT database with 6 different dialects. The stops were detected with an accuracy of 88% (8% substitution, 3% insertion and 1% deletion errors). The classification accuracy was 86%, with 97% voicing accuracy and 90% place of articulation accuracy.
Bibliographic reference. Ali, Ahmed M. Abdelatty / Spiegel, Jan van der / Mueller, Paul (1999): "Automatic detection and classification of stop consonants using an acoustic-phonetic feature-based system", In ICPhS-14, 1709-1712.