Title :
Cross-language adaptation of acoustic models in automatic speech recognition
Author :
Nieuwoudt, C. ; Botha, E.C.
Author_Institution :
Dept. of Electr. Electron. & Comput. Eng., Pretoria Univ., South Africa
Abstract :
The paper presents results for adapting speaker independent acoustic phone models for English speech using relatively small amounts of Afrikaans adaptation data and testing results on a speaker independent Afrikaans test set. Two methods for adaptation are employed; maximum a posteriori probability (MAP) and maximum likelihood linear regression (MLLR) adaptation
Keywords :
maximum likelihood estimation; speech recognition; Afrikaans adaptation data; English speech; MAP; MLLR adaptation; acoustic models; automatic speech recognition; cross-language adaptation; maximum a posteriori probability; maximum likelihood linear regression; speaker independent acoustic phone models; Acoustic testing; Adaptation model; Automatic speech recognition; Automatic testing; Electronic equipment testing; Loudspeakers; Maximum likelihood linear regression; Natural languages; Speech analysis; Speech recognition;
Conference_Titel :
Africon, 1999 IEEE
Conference_Location :
Cape Town
Print_ISBN :
0-7803-5546-6
DOI :
10.1109/AFRCON.1999.820789
