Zimmerman voice analysis: technical

[Jonathan Lighter]

Ron's access is temporarily problematical, so he's invited me to forward
the following from Larry Solan and Anders Eriksson.

Their comments are about the identity of the screamer in the background of
a 911 call at the time of the Martin shooting.  They are not related
(except inferentially) to the interpretation of the inaudible "phrase"
we've been discussing.

Fowarded messages start here:

>Anders is one of the world's foremost experts on voice identification--a
scientifically cautious and conservative phonetician. I thought you might
find it useful to post this on ADS-L, if you agree that it is a point of
view worth sharing with our colleagues.

Larry Solan, by the way, if you don't already know it, is a former
president of the International Association of Forensic Linguists and a dean
in the Brooklyn Law School.

Cheers,
Ron


>>On Apr 4, 2012, at 4:09 AM, Anders Eriksson wrote:

Dear All

I wouldn't attach too much weight to these voice comparisons. I have only
heard a sample of the call that was published by one of the newspapers but
that is at least enough to realise that the sound quality of the questioned
sample is very poor. This is already a severe problem for biometric voice
comparison. The sample is heavily distorted because of the poor microphone
quality of the mobile phone and possibly also clipping due to the fact that
the speaker is shouting.

Shouting in itself also presents a problem. One of the analysts says that
the voice is more likely to be that of a young person. I assume that the
very high voice pitch is one of the reasons for that impression. But we
should then remember that increasing vocal effort has the effect of raising
voice pitch and when screaming it can easily reach a level three times that
of the normal pitch. In a study by me and Hartmut Traunmüller we had
speakers communicate over a range of distances from 0.5 metres to 187.5
metres. For the longest distance they were shouting, but still not quite at
the same level as I perceive in the 911 call in this case. The voice pitch
for male speakers in our case increased from an average of 110 Hz for
normal speech to 274 Hz for shouted speech. In the Zimmerman case the voice
pitch may well have reached levels above 300 Hz which makes reliable voice
comparison very difficult.

Consulting the home pages of the main analyst, I also notice that he
belongs to the "voiceprint" camp. This method was abandoned already during
the late 70-ies in most countries due to its demonstrable lack of
reliability. Some private investigators in the US still use it but
elsewhere it is not recognised at all. The IAFPA does not recognise it as a
valid voice comparison method and analysts using the method are not
accepted as members.

The conclusion based on the above must be that we should not regard these
analyses as reliable evidence in the case.

Anders


>>>On 4 April 2012 00:36, Lawrence Solan <larry.solan at brooklaw.edu> wrote:

> The American media is rightly focused on the Trayvon Martin case, in which
> a young African-American man was killed in Florida by an individual who was
> privately policing the neighborhood and claimed that he shot the young man
> in self-defense.  Screams were heard on the tape of the shooter’s call to
> the police.  The experts say that the screams were not those of the
> shooter, whose voice is on the tape and identifiable.  Here is a link to
> one of many articles in the press describing the analysis:
> http://usnews.msnbc.msn.com/_news/2012/04/01/10963191-trayvon-martin-case-911-call-screams-not-george-zimmermans-2-experts-say
> ****
>
> ** **
>
> Larry****
>
> ** **
>
> Lawrence M. Solan****
>
> Don Forchelli Professor of Law****
>
> Brooklyn Law School****
>
> 250 Joralemon Street****
>
> Brooklyn, NY 11201****
>
> 718-780-0357****
>
> www.brooklaw.edu/lawrence_solan
>

-- 
"If the truth is half as bad as I think it is, you can't handle the truth."

------------------------------------------------------------
The American Dialect Society - http://www.americandialect.org