Hei guys,
How to tackle this problem of pitch detection where I get the pitch Period
detected 22 or 44 or 66 samples but in reality it was always a period of 22
samples. I am using AMDF , does center clipping will help?
Regards,
D.Pal
Pitch detected in multiple of fundamental frequency
Started by ●May 28, 2007
Reply by ●May 28, 20072007-05-28
Debabrata-
> How to tackle this problem of pitch detection where I get the pitch Period
> detected 22 or 44 or 66 samples but in reality it was always a period of
> 22 samples. I am using AMDF , does center clipping will help?
One simple thing to try is keep a history of pitch results. You know pitch cannot
jump instantaneously, so if you get a glitch not consistent with values in frames
before and after, then you could guess that was a false estimate. That assumes you
have at least one buffer of delay.
But that won't help all of the time. To make your pitch detection more reliable,
you're going to need more sophisticated processing than AMDF, which is a simplistic
time domain method.
-Jeff
> How to tackle this problem of pitch detection where I get the pitch Period
> detected 22 or 44 or 66 samples but in reality it was always a period of
> 22 samples. I am using AMDF , does center clipping will help?
One simple thing to try is keep a history of pitch results. You know pitch cannot
jump instantaneously, so if you get a glitch not consistent with values in frames
before and after, then you could guess that was a false estimate. That assumes you
have at least one buffer of delay.
But that won't help all of the time. To make your pitch detection more reliable,
you're going to need more sophisticated processing than AMDF, which is a simplistic
time domain method.
-Jeff
