0
$\begingroup$

I read some recent papers about time-domain speech recognition, and wondering if it steers away from the stationary assumption of FFT. However, some preprocessing operations like pre-emphasis and low-pass filtering still employ FFT. Does it mean waveform-based ASR with these preprocessing still has the FFT assumption of stationarity?

Many thanks!

$\endgroup$
1
$\begingroup$

Pre-emphasis and low-pass filtering can be done completely in the time domain (using IIR or FIR filtering, for instance), thus without an FFT (or with, by fast convolutions filtering methods).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.