14th International Congress of Phonetic Sciences (ICPhS-14)

San Francisco, CA, USA
August 1-7, 1999

Phonemes Classification with Recurrent Neural Networks

Anna Esposito (1,2), R. Ceglia (1)

(1) International Institute for Advanced Scientific Studies (IIASS), Vietri sul Mare (SA), Italy
(2) INFM, Salerno University, Italy

An automatic system, Rasta-PLP and Recurrent Neural Network based, for classifying vowels, fricatives and nasals from TIMIT has been built up. A new philosophy based on the idea to use a short piece of the available signal was implemented in the preprocessing phase. This allowed to save memory space and computational time. On the testing data, the results gave the 98.5% of correct classification for vowels and fricatives, and the 97% of correct classification for nasals. These results are significant considering that they have been obtained using only 30 msec of the signal available.

Full Paper

Bibliographic reference.  Esposito, Anna / Ceglia, R. (1999): "Phonemes classification with recurrent neural networks", In ICPhS-14, 707-710.