Open jollwish opened 10 years ago
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 https://github.com/scoreur/soloAnalysis/issues/1.
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.
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.