In pocketsphinx-0.8, there have a tdt_sc_8k model in "pocketsphinx-0.8\pocketsphinx-0.8\model\hmm\zh".
Is there information about the file?
(1) Such as how many people record the voice?
(2) How many training data ?
(3) How many hours does the model requires of training?
Thanks for your help
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Really thanks a lot, so it looks like that： tdt_sc_8k => Topic Detection and Tracking _ Simplified Chinese _ 8000 sample rate
My question：
If I use tdt_sc_8k to be my model, and using testing data which have 16000 sample rate.
Does it will have a great impact on the recognition rate ?
Because I use testing data which have 16000 sample rate, the accuracy seems like very good.
Thanks for your help again.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello,
In pocketsphinx-0.8, there have a tdt_sc_8k model in "pocketsphinx-0.8\pocketsphinx-0.8\model\hmm\zh".
Is there information about the file?
(1) Such as how many people record the voice?
(2) How many training data ?
(3) How many hours does the model requires of training?
Thanks for your help
You can find it here:
https://catalog.ldc.upenn.edu/LDC2005S11
If you are looking for corpora, it is better to download this:
http://www.openslr.org/18/
Dear sir,
Really thanks a lot, so it looks like that：
tdt_sc_8k => Topic Detection and Tracking _ Simplified Chinese _ 8000 sample rate
My question：
If I use tdt_sc_8k to be my model, and using testing data which have 16000 sample rate.
Does it will have a great impact on the recognition rate ?
Because I use testing data which have 16000 sample rate, the accuracy seems like very good.
Thanks for your help again.
8khz recognition is usually 10% worse than 16khz.
Overall, you need bigger model than tdt, it is not going to be accurate. So you need to collect your data anyway.