I have done some experiments of speaker adaptation for Indian English.
Initially, I have 45 speakers data each has spoken 15 min. Then I have tested
speaker adaptation performance for three new speakers which is 76% 86% 70%.
Now I have added 30 more speakers. Now, for the same speakers for whom I have
done adaptation earlier I am getting 69% 83% 60%. But, the new models are
giving very good results for speaker independent recognition for 45 speakers
63% and now 79%.
The reduction in speaker adaptation is due to variances spread out after
adding more speakers or any other reason. Please give me some suggestion.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Sir,
I have done some experiments of speaker adaptation for Indian English.
Initially, I have 45 speakers data each has spoken 15 min. Then I have tested
speaker adaptation performance for three new speakers which is 76% 86% 70%.
Now I have added 30 more speakers. Now, for the same speakers for whom I have
done adaptation earlier I am getting 69% 83% 60%. But, the new models are
giving very good results for speaker independent recognition for 45 speakers
63% and now 79%.
The reduction in speaker adaptation is due to variances spread out after
adding more speakers or any other reason. Please give me some suggestion.
Yes