First International Conference on Spoken Language Processing (ICSLP 90)

Kobe, Japan
November 18-22, 1990

Improved Broad Phonetic Classification and Segmentation With a Neural Network and a New Auditory Model

Lieven Depuydt, Jean-Pierre Martens, Luc Van Immerseel, Nico Weymaere

Laboratory for Electronics and Metrology, University of Gent, Gent, Belgium

We describe a broad phonetic classification and segmentation algorithm based on auditory modelling, neural networks and dynamic programming. As the basics are outlined in another paper [6], we here focuss on some new elements such as the introduction of auditory model features and durational constraints, and the use of different sets of broad phonetic classes. Furthermore, we have carried out experiments to test the robustness of our algorithm against new speakers and new vocabularies.

Full Paper

Bibliographic reference.  Depuydt, Lieven / Martens, Jean-Pierre / Immerseel, Luc Van / Weymaere, Nico (1990): "Improved broad phonetic classification and segmentation with a neural network and a new auditory model", In ICSLP-1990, 1041-1044.