Speech Recognition Engines

Nested
Different Recording Issues.
User: Dee
Date: 7/10/2020 6:56 am
Views: 3198
Rating: 0

I recorded some interviews using different platforms.  When converted to audio for analysis; some were mono with 16KHz and 69Kbps.  Others were stereo with 32Khz and 128Kbps.

 

Should I covnvert them to the same .wav with mono at 16Khz and 32Bits?

 

Will that affect the learning algorithm?

--- (Edited on 7/10/2020 6:56 am [GMT-0500] by Dee) ---

Re: Different Recording Issues.
User: kmaclean
Date: 7/15/2020 10:38 am
Views: 1481
Rating: 0

>Should I covnvert them to the same .wav with mono at 16Khz and 32Bits?

Sorry, most of my experience has been with HMM-based acoustic models - and for them, yes, they need to be same sampling rate and bits-per-sample

 

--- (Edited on 7/15/2020 11:38 am [GMT-0400] by kmaclean) ---

PreviousNext