[CLAM] Fundamental frequency estimation

Pau Arumi parumi at iua.upf.es
Fri Aug 6 04:03:36 PDT 2004

"Yan Zhou" <yz114 at york.ac.uk> wrote:
> I met a problem while using SMSTools to do the pitchshift transformation. I loaded a 500Hz sinewave (about 1 second time) and set breakpoint to see the fundamental frequency it detects at each frame. The result is 498.204Hz,499.185Hz,499.644Hz,2976.26Hz,2991.05Hz,2991.91Hz....I've tried other sinewaves with different frequencies, and the same thing happened.It seems that it can only detect the first three frame's fundamental correctly, but then goes on weried. While I was using vocal sound instead of sinewave, it works fine. Could anyone tell me why it happens. Thanks!

Hi Yan,

I'm afraid that the pich detection we use in SMSTools is not very flexible. 
Since the sinusoidals tracks detected in a vocal sound and a in a sweep sound are of very different nature, you need to adjust the configuration parameters. 
The best if that you have a look into an example config file.

In this url I've placed a sweep sound file (500Hz-10000Hz), the configuration file sweepConfig.xml (note that you'll have to change the path to the sound files), and a snapshot of the analysis in the SMSTools application:



More information about the clam-users mailing list