Subject: Re: Robust method of fundamental frequency estimation. From: Eckard Blumschein <Eckard.Blumschein@xxxxxxxx> Date: Fri, 2 Feb 2007 10:29:03 +0100 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>Alain de Cheveigne' <Alain.de.Cheveigne@xxxxxxxx> wrote: > pitch perception models > (http://cognition.ens.fr/Audition/adc/pdf/2005_pitch_SHAR.pdf) might > be of use. Thank you, Alain. Admittedly, I did not look into the original posting and the many replies to it, except Dan's, before I replied to him. Meanwhile I read your book chapter quoted above and pondered about the possible reason why it is so difficult to correctly estimate the fundamental frequency at low tones. You seem to still admit a variety of quite different models. I strive for sorting out inappropriate ones. Dan's frequency analysis of A0 on a piano again convincingly demonstrates that at least in this case, pitch cannot be perceived from any frequency out of the spectral pattern. Obviously, the fundamental frequency is to be seen as the distances between the spectral peaks. I conclude, there are at least two lower frequency limitations: Audibility of pure tones is known to depend on SPL. I expect an additional lower frequency limit for recognition of fundamental frequency to depend on the critical bandwidth. So you might wonder why cochlear frequency analysis is not more acute? Well this is understandable to those like me who have minimal knowledge in signal processing: In order to reach high accuracy in subsequent time domain, one has to have enough bandwidth. Please check and, if they turn out correct, then consider these arguments within the next edition of the book. Sincerely, Eckard Blumschein