Advanced Algorithms and Architectures for Speech Understanding
This book is intended to give an overview of the major results achieved in the field of natural speech understanding inside ESPRIT Project P. 26, "Advanced Algorithms and Architectures for Speech and Image Processing". The project began as a Pilot Project in the early stage of Phase 1 of the ESPRIT Program launched by the Commission of the European Communities. After one year, in the light of the preliminary results that were obtained, it was confirmed for its 5-year duration. Even though the activities were carried out for both speech and image understand ing we preferred to focus the treatment of the book on the first area which crystallized mainly around the CSELT team, with the valuable cooperation of AEG, Thomson-CSF, and Politecnico di Torino. Due to the work of the five years of the project, the Consortium was able to develop an actual and complete understanding system that goes from a continuously spoken natural language sentence to its meaning and the consequent access to a database. When we started in 1983 we had some expertise in small-vocabulary syntax-driven connected-word speech recognition using Hidden Markov Models, in written natural lan guage understanding, and in hardware design mainly based upon bit-slice microprocessors.
Cosa dicono le persone - Scrivi una recensione
Nessuna recensione trovata nei soliti posti.
Altre edizioni - Visualizza tutto
Advanced algorithms and architectures for speech understanding, Volume 1
Visualizzazione frammento - 1990
allocated AND-OR trees application architecture average number beam search beam search threshold biport memory buffer candidate words CDI1 CDI2 cepstral classification codebook computational conceptual graphs configuration constraints correct word corresponding CSELT deduction instance defined dependency rule DHMM digital signal processor diphones DSP board DSP1 DSP2 dynamic programming efficiency evaluation fact feature extraction FIFO Figure firmware frame goal hence Hidden Markov Models ICASSP IEEE implementation inclusion rate input isolated words K-LEAF knowledge language large vocabulary lattice lexical access lexicon master micro-segment module number of candidate obtained operator OR-parallel output parallel parameters parser parsing performed phonetic classes phonetic segmentation phrase hypothesis problem Proc processor quality factor representation represented samples score selected sentence sequence sequential solution speaker speech recognition speech understanding structure sub-word units subgoal subwords symbol synchronization Table tree utterance vector quantization verification Viterbi VME bus word hypotheses