György E. Révész
Theoretical Computer Science
The large-vocabulary speech recognition algorithms for implementation of next-generation speech recognition applications are discussed. The lexicon, language model, and acoustic model component are used to construct a hypothesis model used for the interpretation of speech sample. The discussion shows that the word error rate for spontaneous speech sources remain relatively high and the advanced speech recognition (ASR) system remains sensitive to background noise.
György E. Révész
Theoretical Computer Science
Khaled A.S. Abdel-Ghaffar
IEEE Trans. Inf. Theory
Pradip Bose
VTS 1998
Apostol Natsev, Alexander Haubold, et al.
MMSP 2007