"WhistleMark"-absolute measure of whistle quality

[A-Musing]

Ooops....is this data right? The Perfect Whistle is Mariah Carey?

[eskin]

We're still working with the primary investor to figure out how to publish the results. They want some sort of website subscription model with paid memberships and want to expand beyond just whistles to flutes, Uilleann pipes, and other folk instruments beyond the Irish tradition. I don't think that will fly, but with the rumors of a certain Japanese instrument company possibly about to hit the scene with a whole range of premium whistles and flutes (even talk about pipes), we're getting a lot of pressure to hold off releasing any results until the announcements at the NAMM show in Anaheim later this month. I hate this part of the business.

[OBrien]

My other leg is feeling it now.

[Wanderer]

I think the degree to which the testing agrees with our preconceptions and experiences with whistles will determine how we respectively gauge the "Eskin Test"'s validity.

Happy New Year to all.

Philo

I think that's when I feel the most satisfied about one of my whistle reviews: When a few people reply and say "that's exactly how I'd describe my <insert brandname here>"

[Denny]

My other leg is feeling it now.

you might enjoy being taller.

downsides: you will need to watch you head & you will have trouble finding pants

[Loren]

And April still so far off......

Loren.

[Denny]

have ya tried thinkin' of it as a windup?

[PhilO]

Well, I for one will sit here through the new year awaiting the exciting results...the Japanese company, is it Honda or Kawazaki, so we can play really fast...ooh, I'm feeling something a wee bit in the legs too....

Philo

[MTGuru]

Michael and I have actually been working on this idea for a while. In order to address anticipated concerns about the clinical nature of the WhistleMark benchmarks, he's asked me to contribute my modest expertise in computational linguistics as former director of a major language software company to help "humanize" the results.

For example, software for speaker-independent voice recognition makes use of hidden Markov modeling and Bayesian probability for predictive analysis of the morphophonemic segmentation of a non-discrete spoken stream. This same approach can be used to algorithmically isolate time constrained fluctuations of whistle tone introduced by both player-specific performance characteristics and the whistle's own note to note transition contours. Even a medium-sized corpus of recorded samples of typical ITM melodies can produce a statistical baseline for weighting the results.

Psychoacoustic studies have also shown that the subjective perception of tonal quality is highly conditioned by the native language of the hearer. For example, an infant's pre-linguistic babbling and tone perception during the first year of life tends to reflect the tonal and suprasegmental characteristics of the mother's or caregiver's language. And speakers of tonal languages like Mandarin, Twi, or Tzotzil Mayan evaluate musical pitch in terms of their languages' tonal inventory. By correlating the raw WhistleMark output with a sample of multilingual listener and infant responses, it should be possible to generate whistle profiles with greater cross-cultural validity.

Of course, we realize that, as with language translation, no software is perfect. But an application to automatically identify and evaluate whistle tone could be useful, or at least interesting, to makers, players and experimenters alike.

[Denny]

aw, I love it when you talk like that!

If youse ain't careful with that kinda talk this thread is gonna get moved the the PROCTology Forum

[ChaplainBlake]

Ahh, so the goal is to discern the common denominator in quality tone and then mass produce it (that which appeals to the largest range of people). I think this has been done in every other musical field in the past, except human ears were used to determine a consensus, rather than computers (hence, the never ending stream of Stradivarius copies and Fender Stratocaster variants).
Yamaha is probably the culprit then, as they do a great job of making instruments of that sort. Hey, it might help further diversify or multiply the audience for this music, so it cannot be all bad.

[eskin]

Here's a demo of an early version of the software. This video only shows the three-note harmonic baseline static analysis test:

http://www.youtube.com/watch?v=F6xNqEXjQVY

The final UI will be more or less similar, trying to keep a Web 2.0 sort of feel.

If we can get the CPU requirements low enough, a limited version could run on the iPhone once Apple opens up the platform later this month. Probably also on the newer Blackberrys and Treos, would need at least a 400 MHz ARM9 core and audio recording capability.

With the mobile platform, tests could be done in the field, for example just by making a random recording during a session, with the captures plus some additional metadata (whistle make, material, etc.) uploaded to the main site for full analysis when the phone is docked. Of course, a full evaluation would require both melodic and static testing. The main site can aggregate the reports from various makers, give suggestions along the lines of "if you like this whistle, you'd probably also like ..." based on knowing what instruments people already have.

[Tommy]

OK, so where are the comments of Peter Laban?

[walrii]

If I understood what eskin and MTGuru are talking about, I would be VERY impressed.

[wvwhistler]

If this was a truly practical application and the definition of the perfect whistle was achieved and then everyone used that definition for making whistles all we would have is a world of whistle makers creating the exact same thing. The only thing that could possibly change would be appearance and price. The thought of eating hamburgers for every meal doesn't sit well for me.