I am using hub4wsj_sc_8k generic model to do some speech recognition. But here is a question. Do I need downsample all of my input files to 8k sample rate? (The original is 16k)
Thank you.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
It's not really necessary. There are max and min frequency in feat.params files. mel filters are built and used only for that space, so it should be ok. It is suggested to use one of default en-us models though.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi, dear all,
I am using hub4wsj_sc_8k generic model to do some speech recognition. But here is a question. Do I need downsample all of my input files to 8k sample rate? (The original is 16k)
Thank you.
It's not really necessary. There are max and min frequency in feat.params files. mel filters are built and used only for that space, so it should be ok. It is suggested to use one of default en-us models though.