A Self-Organizing-Neural-Network(SONN) architecture is used, in cooperation with Viterbi decoding and One-Pass Level-Building(LB) tecnique for automatic alignment of incoming speech signals with their corresponding indipendently specified string of phonetic/phonemic symbols. Results of the application of the label alignment procedure to EUROMO Italian continuous speech material are presented.
A Sonn-Based Architecture for Automatic Speech Segmentation and Alignment
Contributo in atti di convegno
Proceedings Primo Workshop Italiano su "Le reti neuroniche per il trattamento del parlato", pp. 18–29, Roma, Italia, 8-9 November,1990