In our app we are using PocketSphinx for speech recognition, with Language model generated from Online tool and "hub4wsj_sc_8k" acoustic model.
Recognition is very very good if we use audio generated from TTS for US accent. But we are facing issues with Indian accent. Recognition is very poor with our speech input.
Please suggest a better way to improve the recognition.
Thanks
Swapna
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
In our app we are using PocketSphinx for speech recognition, with Language model generated from Online tool and "hub4wsj_sc_8k" acoustic model.
Recognition is very very good if we use audio generated from TTS for US accent. But we are facing issues with Indian accent. Recognition is very poor with our speech input.
Please suggest a better way to improve the recognition.
Thanks
Swapna
Train Indian English acoustic model
Is traing is the only option available to solve my problem?
How good will acoustic adaptation works in this scenario.
Thanks
Swapna
Yes
It will require the same amount of data as new training
Thanks for the quick response.
I too need help in Indian English accent recognition. Have you successfully trained your model? if yes then how?
Hi Swapna S and Nickolay V. Shmyrev ,
You have any Trained Indian English acoustic model data files . Please share with me ..
Or else how to include the Trained Indian English acoustic model in android app. Any examples you have share the link please .
I was tried the default english model but i got only 40 to 50 % of accuracy .
Are I pasted my Trained Indian English acoustic model data in Asset folder ?
We are going to release Indian English soon, contact me in private nshmyrev@gmail.com if you want to beta test.
@Nickolay please send it me too