Title :
Two-Stage Speaker Adaptation of Hybrid Tied-Posterior Acoustic Models
Author :
Stadermann, Jan ; Rigoll, Gerhard
Author_Institution :
Inst. for Human-Machine Commun., Technische Univ. Munchen, Germany
fDate :
March 18-23, 2005
Keywords :
adaptive signal processing; gradient methods; hidden Markov models; maximum likelihood estimation; neural nets; speech recognition; HMM mixture coefficients; automatic speech recognition systems; gradient based adaptation strategy; hybrid neural network/HMM systems; hybrid tied-posterior acoustic models; neural network adaptation; neuron selection method; scaled likelihood maximization; speaker independent ASR; two-stage speaker adaptation; word error rate; Automatic speech recognition; Context modeling; Error analysis; Hidden Markov models; Loudspeakers; Man machine systems; Maximum likelihood linear regression; Neural networks; Space technology; Technological innovation;
Conference_Titel :
Acoustics, Speech, and Signal Processing, 2005. Proceedings. (ICASSP '05). IEEE International Conference on
Print_ISBN :
0-7803-8874-7
DOI :
10.1109/ICASSP.2005.1415279
