A phonetician's comment on Dmitry Terez's pitch determinator algorithm

[Dinoj Surendran <dinoj@xxxxxxxxxxxxxxx>]

Dear list,

One of the major users of a pitch tracking algorithm would be experimental
phoneticians, particularly those whose investigations include prosody and
tones. I told one of them, Yi Xu (xuyi@uchicago.edu) about the new
algorithm and asked him what he thought. He's given me permission to post
his reply to the list :

>On Wed, 23 Jul 2003, Yi Xu wrote:
>
>> Dinoj,
>>
> > I tried the demo and it even skipped the creaky part in the example
> sentence he gave. I also tried a few of our own wave files. Pitch of the
> voiced sections are frequently missed. While there may be a gem hidden
> in the algorithm, the author apparently are not really familiar with all
> the hurdles a real-life pitch extraction algorithm has to be able to
> handle.
>
> Basically, a pitch extraction algorithm with superiority claims should
> at least show that it can handle all the well-known difficulties in
> pitch extraction, including pitch halfing, pitch doubling, voicing
> detection, glottalization, etc. I have yet to see any pitch extraction
> algorithm handle these problems better than the semi-manual procedure I
> have been using for my own research. I look forward to the day when an
> algorithm can meet my stringent but very reasonable standard. If the
> author truly believes that the algorithm has the potential to solve all
> these problems, I will be happy to show him some concrete examples. But
> as a first step, he should at least give us a flawless demo :-).
>
> Yi

Hopefully the algorithm can be modified to deal with the problems laid out
above.

Dinoj Surendran
Graduate Student
Department of Computer Science
University of Chicago
http://www.cs.uchicago.edu/~dinoj