15th International Congress of Phonetic Sciences (ICPhS-15)

Barcelona, Spain
August 3-9, 2003

An Event-Based Acoustic-Phonetic Approach for Speech Segmentation and E-Set Recognition

Amit Juneja, Carol Y. Espy-Wilson

University of Maryland, USA

In this paper, we discuss an automatic event-based recognition system (EBS) that is based on phonetic feature theory and acoustic phonetics.
   First, acoustic events related to the manner phonetic features are extracted from the speech signal. Second, based on the manner acoustic events, information related to the place phonetic features and voicing are extracted. Most recently, we focused on place and voicing information needed to distinguish among the obstruents /b/, /d/, /p/ /t/, /s/, /z/ and /jh/. Using the E-set utterances from the TI46 test database, EBS achieved place accuracy of 96.35% and voicing accuracy of 96.53% for the above mentioned obstruents. The recognition accuracy of EBS is presented at its different stages of classification. The applications of the system apart from phoneme recognition are also discussed.

Full Paper

Bibliographic reference.  Juneja, Amit / Espy-Wilson, Carol Y. (2003): "An event-based acoustic-phonetic approach for speech segmentation and e-set recognition", In ICPhS-15, 1333-1336.