tape identification . . .

[JES]

Okay, this is a thread on Osama bin Laden that's actually meant to be about recording (as opposed to politics or laughs). Many of you are no doubt aware of a new tape that the U.S. claims contains Bin Laden's voice. Thing is, it's a bad cassette recording of a speech over a telephone line. My question is: how do you think the NSA is analyzing this tape? All the NYTimes says is they're using "digital" techniques. Well, woo-hoo! I'm curious as to how it's actually done. If you know any good publications on voice identification through a DAW, or have done some of it yourself, I'd be really interested to know about the art and science of it. Thanks. --JES

[MattC]

Just shooting off the wall here, but: Every instrument has certain harmonic overtones; this is the reason why a piano's middle C sounds different than that of a bass. I'm sure that a lot of those "digital techniques" involve spectral analyzation of the voice on the tape and a comparison to known samples of Bin Laden's voice.

[GT40sc]

NPR's "All Things Considered" had a good story about this on Wednesday afternoon (11-13) Not very technical, but informative. They inteviewed Steve Kane of Forensic Tape Analysis Inc., Lake Geneva, Wisconsin. The story that followed was an interview with an Egyptian political scientist here in the U.S., who felt that the tape did not contain Bin Ladin's voice. npr.org hope this helps,

[D_dup3]

Maybe they could send it to one of those labs like in TV shows (CSI; Agency: etc.)! :thu: They all seem to be able to do [i]the most unbelievable[/i] things with their computers, etc. :rolleyes:

[Wewus432]

It seems like a very simple process to me. They have many known copies of his voice. You would isolate specific words, phrases, syllables, etc. of the known audio in a program like Sound Forge, Cool Edit or WaveLab, then you isolate the same words, phrases, syllabes, etc. in the new audio and LOOK for matches. You would also have to look for digital splices because they could have made the new message by using the same software and cutting and pasting the words together.

[JES]

[quote]Originally posted by GT40sc: [b]NPR's "All Things Considered" had a good story about this on Wednesday afternoon (11-13) Not very technical, but informative. They inteviewed Steve Kane of Forensic Tape Analysis Inc., Lake Geneva, Wisconsin. The story that followed was an interview with an Egyptian political scientist here in the U.S., who felt that the tape did not contain Bin Ladin's voice. npr.org hope this helps,[/b][/quote]Thanks -- this looks like a great lead. Dunno if you all remember this, but back in the 1980s there was a tape circulating purporting to be the voice of "Elvis" which had gone through an identification process. Of course, they found some voiceprint expert to say that it was him. Not saying that the voice on the bin Laden tape isn't him (how would I know), just wondering how/whether the science of it has changed significantly since the 1980s. Best, --JES

[MattC]

Analog vs. digital: the science of sound has changed forever with the emergence of the omnipresent digital realm. You probably could buy better analyzing gear nowadays for the value of your car than they had in the mid '80's.

[D_dup3]

Contrary to what we might believe from science fiction & TV melodramas, some things are not possible...& it may be possible for sounds to be so altered/garbled as to render them beyond perfect identification. The incotrovertability of science is contradicted by history. In any case, what would identifying the tape's speaker prove?

[Wewus432]

Have you guys ever done any digital editing with a program like Cool Edit? It's unbelievable. Things have changed A LOT since the eighties. You can [b]look at and listen to[/b] sounds down to the [b]bit[/b] . You can amplify, reverse, turn upside down, transpose, slow down, and analyse for frequency range and amplitude ANY sound. And that's just a start. The great thing is you have a visual representation of the waveform and every change you make to it. That's why I say it should be an easy process to identify the speaker on a tape if you already have a known copy of the voice. Just a matter of matching the unique visual representation of the waveform in the known voice to the waveform of the unknown voice.