f0_test is raw

[jollwish]

The present f0_test is based on auto-correlation and is vulnerable to real-life noised audio. Something further has to be done such as denoising, combination with other f0 analysis algorithm, or Baysian method usage.

[chenxiaoqino]

Is it possible to add a denoise filter beforehand to solve this problem
with minimum efforts? Or, we can exploit the rules between frequencies of
regular scores, (2^(1/16)) times between each other, to distinct them from
irregular noise.

Yours sincerely, Danny Chen
计科30 陈晓奇
about.me http://about.me/cxqn

On Sat, Aug 16, 2014 at 1:46 AM, jollwish notifications@github.com wrote:

The present f0_test is based on auto-correlation and is vulnerable to
real-life noised audio. Something further has to be done such as denoising,
combination with other f0 analysis algorithm, or Baysian method usage.

—
Reply to this email directly or view it on GitHub
#1.

[jollwish]

Yes and I've tried several denoising algorithms. But one problem still remains that it's not so easy to distinguish f0,2f0, or 1/2f0. I think Bayesian method can solve this and Wang Kangning is working on that.