How do they identify (or try to) BinLaden's (et al) voice?

[[email protected]]

When you read the news, stories about how they're trying to verify
whether or not a certain tape played Bin Laden's voice (or Saddam's),
what programs are they using? I imagine a control room akin to some
sort of NASA labratory but I wonder if in fact they're simply using a
mix of commercially available plug-ins--the lastest speech software,
for example. Anyway, just curious if anybody knew as I sit here
pondering how speech identification might work behind the big closed
doors.

[dt king]

> wrote in message
om...
> When you read the news, stories about how they're trying to verify
> whether or not a certain tape played Bin Laden's voice (or Saddam's),
> what programs are they using? I imagine a control room akin to some
> sort of NASA labratory but I wonder if in fact they're simply using a
> mix of commercially available plug-ins--the lastest speech software,
> for example. Anyway, just curious if anybody knew as I sit here
> pondering how speech identification might work behind the big closed
> doors.

A CIA team studies the tape for verification. As they're just down the
hall from the team who produces it; just a matter of comparing it to the
script.

dtk

[Scott Dorsey]

In article >,
> wrote:
>When you read the news, stories about how they're trying to verify
>whether or not a certain tape played Bin Laden's voice (or Saddam's),
>what programs are they using? I imagine a control room akin to some
>sort of NASA labratory but I wonder if in fact they're simply using a
>mix of commercially available plug-ins--the lastest speech software,
>for example. Anyway, just curious if anybody knew as I sit here
>pondering how speech identification might work behind the big closed
>doors.

I don't know how it's done today, but a decade ago it was mostly a matter
of printing out speech spectrograms and matching the patterns of different
sounds by eye.

It wasn't particularly easy to do to a high confidence level with English
language material; I'd hate to try and do it with material in a language I
don't know or have a good feel for the rhythms of.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Ricky W. Hunt]

"Scott Dorsey" > wrote in message
...
> In article >,
> > wrote:
> It wasn't particularly easy to do to a high confidence level with English
> language material; I'd hate to try and do it with material in a language I
> don't know or have a good feel for the rhythms of.

Don't they still rely a great deal on the good old human ears of people who
knew him?

[Ben Bradley]

In rec.audio.pro, (Scott Dorsey) wrote:

>In article >,
> > wrote:
>>When you read the news, stories about how they're trying to verify
>>whether or not a certain tape played Bin Laden's voice (or Saddam's),
>>what programs are they using? I imagine a control room akin to some
>>sort of NASA labratory but I wonder if in fact they're simply using a
>>mix of commercially available plug-ins--the lastest speech software,
>>for example. Anyway, just curious if anybody knew as I sit here
>>pondering how speech identification might work behind the big closed
>>doors.
>
>I don't know how it's done today, but a decade ago it was mostly a matter
>of printing out speech spectrograms and matching the patterns of different
>sounds by eye.

There's also the words and phrases used on the tape, in comparison
to the text of known speeches and/or writings. When previously unknown
writings, allegedly by a famous writer, are found, things such as word
and phrase frequency are compared to the known body of work to see if
the questioned text is statistically similar with the known texts.
The longer the text on the tape (and the more known text available),
the more accurate the determination is likely to be.
This of course isn't about identifying a voice (and also has
nothing to do with audio), but it gives extra info in identifying a
person, which is the idea anyway.

>It wasn't particularly easy to do to a high confidence level with English
>language material; I'd hate to try and do it with material in a language I
>don't know or have a good feel for the rhythms of.
>--scott
>--
>"C'est un Nagra. C'est suisse, et tres, tres precis."

[It's me!]

On 4 Sep 2003 12:00:39 -0700, wrote:

>When you read the news, stories about how they're trying to verify
>whether or not a certain tape played Bin Laden's voice (or Saddam's),
>what programs are they using? I imagine a control room akin to some
>sort of NASA labratory but I wonder if in fact they're simply using a
>mix of commercially available plug-ins--the lastest speech software,
>for example. Anyway, just curious if anybody knew as I sit here
>pondering how speech identification might work behind the big closed
>doors.

You've heard how Bin Laden doesn't believe in toilet paper and uses
his fingers instead... well, they smell the microphone; if it smells
like dookie, it was him.

[WillStG]

<< (Ben Bradley) >>
<< There's also the words and phrases used on the tape, in comparison
to the text of known speeches and/or writings. When previously unknown
writings, allegedly by a famous writer, are found, things such as word
and phrase frequency are compared to the known body of work to see if
the questioned text is statistically similar with the known texts.
The longer the text on the tape (and the more known text available),
the more accurate the determination is likely to be.
This of course isn't about identifying a voice (and also has
nothing to do with audio), but it gives extra info in identifying a
person, which is the idea anyway. >>

They are also examining the material forensically to try to determine if
it is edited material. As I understand it, many of these tapes are
intentionally several generations down from the original recordings. When the
source material is incredibly noisy and hissy, it makes it very difficult to
determine if a tape is a composite or not. The terrorists also wish to mask
any background noises that might provide clues to our intelligence services as
to the location of the recording, this also makes it difficult to determine if
it was simply an edited product of old recordings.


Will Miho
NY Music & TV Audio Guy
Fox And Friends/Fox News
"The large print giveth and the small print taketh away..." Tom Waits