Vocal Fingerprinting

May 12, 2004 7:09 PM

For my high school's senior research project, I'm considering doing research on vocal fingerprinting.
I'm looking for a free repository of samples of a number of people speaking the same set of things. Does this exist?
In a similar vein, my lab director is trying unsuccessfully to find suppliers for the Motorola 68008. We'd like a few dozen, but we'd be happy to get our hands on any. Anybody know where to get these? [more inside]

The George Mason University Accent Archive came to mind, but I don't think it's what I'm looking for -- there's only one sample per person, so the fingerprinting wouldn't be very interesting. I could probably gather my own library of samples from unsuspecting fellow students and teachers, but that would involve time and effort. (The horror, eh?)

And the M68008 -- the program takes a very hands-on approach, and we breadboard our systems from scratch. We have working implementations of the Motorola 6802, but it's just not powerful enough for real-time FFT's or DSP applications. The 68k8 provides an increase in power without excessively boosting complexity -- like the 6802, it has an 8-bit wide data bus, which lets us get away with less bus wiring. Any recommendations for similar devices (available in DIP packages, 40-50 pins) would be tremendously appreciated.

posted by mote to Technology (5 answers total)

Hey mote! I went to tj.... eight years ago. < /derail>
posted by croutonsupafreak at 11:06 PM on May 12, 2004

There is the speech accent archive, that examines the accented speech of people from around the world reading the same sample paragraph. The current 337 samples also includes phonetic generalizations.
posted by KathyK at 6:47 AM on May 13, 2004

Your choice of the 68008, I assume, is being made in order to simplify the design of the product. The 68K8 was Motorola's attempt at making a chip with the power of a 68K but with a slightly simplified addressing scheme (8 bit data bus). The problem with this is that the chip is really a 16/32 bit data bus chip, which means that in order to fetch a long word, it has to hit the data bus 4 times. Since the chip has no on-board instruction cache, your execution time goes in the toilet. People were getting better performance by cranking up the clock rate on a 6809 or Z-80 type chip instead, so it's no surprise that you can't find it.

If you want to stay in the 68K family, you might consider instead the Dragonball family of processors. They were (are?) used in Palm and Handspring devices and the specs look like they are made for hooking up to things so getting it wired up should be no big deal. I think your biggest concern will be that the chip probably only comes in a TQFP form factor, which is surface mount - but I think you could cope with this by putting it on its own PCB which is sized to the DIP form factor.
posted by plinth at 10:51 AM on May 13, 2004

If you don't have a Digi-Key catalog, get one. They're willing to sell in small quantities to non-corporations, and their catalog is huge enough to be useful as a guide to what's available out there.

(plinth probably has the right idea about getting something in a surface mount package and making up a carrier board for it --- this would let you move to a more recent, and therefore easier-to-find, 68k variant. Or you could build your own 68008... )

Also, the Usenet newsgroups comp.arch.embedded and sci.electronics.* are alive and well and might be able to help you.
posted by hattifattener at 11:45 AM on May 13, 2004

Response by poster: neustile: Awesome! Is there an easy way to get it that I'm overlooking, or should I just start emailing people and pleading my case?

Thanks for the warning, plinth. I'll have to pass that along... and thanks for the Dragonball link. The sizable data bus sounds particularly nice.

Thanks hattifattener for the Usenet pointers! And one of my friends was mocking PGA's earlier... I'll have to show him that, heh.
posted by mote at 6:25 PM on May 13, 2004