Proceedings of the 1998 IEEE International Conference on Acoustics, Speech, and Signal Processing: May 12-15, 1998, Washington State Conventon and Trade Center, Seattle, WA (USA)IEEE Service Center, 1998 - Electro-acoustics |
From inside the book
Results 1-3 of 83
Page 76
... Adaptation classes 11 12 13 6. CONCLUSION We have presented an adaptation algorithm which approximates the theoretically optimal EMAP algorithm , and requires far less computation . The basic idea is to use a simplified iterative optimi ...
... Adaptation classes 11 12 13 6. CONCLUSION We have presented an adaptation algorithm which approximates the theoretically optimal EMAP algorithm , and requires far less computation . The basic idea is to use a simplified iterative optimi ...
Page 86
... Adaptation Approach As in the speaker - independent case , the target of ML based speaker adaptation is to modify the model parameters so that the likelihood of the adaptation utterances is maximized . Two widely known ML based adaptation ...
... Adaptation Approach As in the speaker - independent case , the target of ML based speaker adaptation is to modify the model parameters so that the likelihood of the adaptation utterances is maximized . Two widely known ML based adaptation ...
Page 115
... Adaptation Each GMM is adapted individually using sub - word data ac- quired from the blind segmentation . The remainder of this section describes the adaptation of a single sub - word GMM since the process is identical for each sub ...
... Adaptation Each GMM is adapted individually using sub - word data ac- quired from the blind segmentation . The remainder of this section describes the adaptation of a single sub - word GMM since the process is identical for each sub ...
Contents
Intelligent Query and Browsing Information Retrieval IQBIR Agent II1173 | xxxi |
LOWBIT RATE SPEECH CODING | xxxii |
SPEECH ANALYSIS AND SYNTHESIS | xxxiii |
Copyright | |
64 other sections not shown
Common terms and phrases
acoustic models adaptation algorithm analysis applications approach Audio average beamforming bigram bit rate CELP cepstral cepstrum channel classification clean speech codebook codec coder coefficients computed context continuous speech corpus correlation database decoder density digit discriminative distortion encoding enhancement estimation evaluation experiments feature vector Figure filter frame frequency function Gaussian Hidden Markov Models hybrid ICASSP improvement input iteration kbit/s language model likelihood method microphone mixture MP-CELP n-gram neural networks noise obtained optimal output parameters performance phoneme pitch probability Proc proposed recognizer reduced robust samples score segment sentences sequence Signal Processing speaker speaker verification spectral speech coding speech data speech recognition system speech signal statistical string subword synthesis Table task techniques Technology tion training data transform triphone University utterances variable variance vector quantization verification Viterbi Viterbi search vocabulary voiced weight word error rate