ISCA Archive ICSLP 1992
ISCA Archive ICSLP 1992

Real-time speaker-independent large-vocabulary CDHMM-based continuous telephonic speech recognizer

Alberto Ciaramella, Davide Clementino, Roberto Pacifici

We describe a real-time continuous-speech recognition system for large vocabulary, speaker independent telephone accessed applications implemented on a multiple floating-point DSP environment. Both Continuous and Discrete Hidden Markov Models have been implemented; we summarize the performance of both cases and we detail the structure of the multi-DSP implementation.