Dear Nickolay,
so thankyou for your help before :)
i have done some basic testing on my Vietnamese speech recognition system, but
i found that the accuracy of my system was degrate too much in noise
environment. So, i wanna to implement and integrate a speech enhancement
module in to pocketsphinx. I need some advice from you.
i consider 3 tradition method : OMLSA, MMSE and spectral subtraction
do you have any idea ?
and one more is should i implement it in pocketsphinx or sphinxbase and in
what stage ?
thanks in advance.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
thanks for your idea, but i have a small question : does CMU has any project
about MVDR noise cancellation and did you try to integrate it into sphinx ?
and have any student or professor study in this method, can i have their
contact
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Nickolay,
so thankyou for your help before :)
i have done some basic testing on my Vietnamese speech recognition system, but
i found that the accuracy of my system was degrate too much in noise
environment. So, i wanna to implement and integrate a speech enhancement
module in to pocketsphinx. I need some advice from you.
i consider 3 tradition method : OMLSA, MMSE and spectral subtraction
do you have any idea ?
and one more is should i implement it in pocketsphinx or sphinxbase and in
what stage ?
thanks in advance.
The best method recently promoted is MVDR noise cancellation
A SINGLE-CHANNEL NOISE REDUCTION MVDR FILTER
Authors: Jacob Benesty, INRS, Canada; Yiteng (Arden) Huang, WeVoice Inc.,
United States
http://mirlab.org/conference_papers/International_Conference/ICASSP%202011/pd
fs/0000273.pdf
thanks for your idea, but i have a small question : does CMU has any project
about MVDR noise cancellation and did you try to integrate it into sphinx ?
and have any student or professor study in this method, can i have their
contact
I'm not aware about anything like that.