Sixth European Conference on Speech Communication and Technology

Budapest, Hungary
September 5-9, 1999

Speaker Normalization for Audio-Visual Articulation Training

Marcel Ogner, Zdravko Kacic

Faculty of Electrical Engineering and Computer Science, University of Maribor, Slovenia

The paper describes formant based speaker zation method suitable for speech visualization and articulation training systems. The method estimates the error function obtained from speaker formant characteristics for a given vowel. Estimated error function gives information for critical band filter shifting on mel-warped frequency scale. The paper also describes accurate technique for formant tracking.

Full Paper (PDF)

Bibliographic reference.  Ogner, Marcel / Kacic, Zdravko (1999): "Speaker normalization for audio-visual articulation training", In EUROSPEECH'99, 579-582.