Voice ID experts

[RickyJim]

There is nothing in the two reports to indicate the authors had any basis for comparing screams on a cell phone with whatever other samples they had.  In fact one report admits that they hadn't made any control studies and wishes they had more time and money to do them.   Neither report represents standard methods that have been accepted in courts to do voice comparison.  BDLR has some nerve to suggest that a Frye hearing is not necessary to discuss the validity of the "science" used.

[Cylinder]

Riech made a huge error in the claim that he could extract a voice superimposed over a gunshot.

[cboldt]

Alan Reich has been on the State's side since long before he got Mr. Martin's exemplars. In fact, his document is pretty much a copy/paste of what he wrote nearly 1 year ago:

http://www.washingtonpost.com/national/trayvon-martin-case-911-call-two-experts-reach-two-much-different-conclusions/2012/05/19/gIQAtuapbU_story.html

Heh.  I figured the state was smart enough to avoid experts (like Own and Primeau) who had made public pronouncements.  I guess the state isn't that smart.  Reich is just sticking to his previous in order to protect his own reputation.

His argument essentially starts with the conclusion, and then makes findings.  The person making the controlled sounds is not the screamer, and since Zimmerman won the fight (had the upper hand, the fight was over by the time the 911 call started!), his voice is the one that is more controlled.

[RickyJim]

From page 5 of the H and H report,

Both the "known" and the "unknown" samples were band-pass filtered with both the high pass and low pass filtering cutoffs set outside the speech range.

  I am wondering if those cutoffs clipped the scream range especially at the high end.

[cboldt]

Just a quick thought about whether or not the experts are to be subjected to a Frye hearing.  As with most any issue in the law, a fundamental pivot point is defining the question or framing the issue.  If the question is "is spectrographic analysis accepted in the scientific community?" the answer is "Yes."  But, if the question is "is there a scientifically accepted method to perform voice ID from a sample of a person under stress, with the sample voice being outdoors in a reverberating environment, communicated through (50) feet of air, to a phone, and recorded on 8 kHz 911 digital equipment (and the sample is contaminated with other, overlapping sounds)?"  The answer is no - see battling experts.  If there was a scientifically accepted method for this, the experts would agree on the ultimate answer.

[cboldt]

Another quick thought.  I don't know if interpretation of equivocal sounds into words and phrases passes the Frye test.  The voice analysis that I am aware of uses audio-spectral analysis to associate spoken words with the person who spoke them.  I am not aware of audio-spectral analysis being reliably used to disambiguate utterances.

Bernardo, you lose Reich (if the law is followed) at least as to word identification, the jury will decide themselves what words they hear.

[RickyJim]

Just a quick thought about whether or not the experts are to be subjected to a Frye hearing.  As with most any issue in the law, a fundamental pivot point is defining the question or framing the issue.  If the question is "is spectrographic analysis accepted in the scientific community?" the answer is "Yes."  But, if the question is "is there a scientifically accepted method to perform voice ID from a sample of a person under stress, with the sample voice being outdoors in a reverberating environment, communicated through (50) feet of air, to a phone, and recorded on 8 kHz 911 digital equipment (and the sample is contaminated with other, overlapping sounds)?"  The answer is no - see battling experts.  If there was a scientifically accepted method for this, the experts would agree on the ultimate answer.

Standard courtroom accepted science is to use spectral analysis to compare 10-20 snippets of pairs of syllables taken from two samples made under (almost) identical conditions.  Anything that departs essentially from that model should be subject to a Frye hearing.

[cboldt]

Standard courtroom accepted science is to use spectral analysis to compare 10-20 snippets of pairs of syllables taken from two samples made under (almost) identical conditions.  Anything that departs essentially from that model should be subject to a Frye hearing.

Bernardo is arguing that the scope of Frye hearing is limited to identification of the technology, in this case, spectral analysis.  The point that both you and I are making is that the scope of Frye hearing necessarily involves identification of more than just the technology, it also has to look at the relationship between the technology and the evidence.  That is an issue in any audio-spectral case, and so there should be a hearing.

[RickyJim]

What is missing is the original request from ASA Richard W. Mantei to Hollien and Harnsberger to do the testing.  I assume the defense is allowed to see it.  It is unclear what they were asked to do.  Their report shows that they tested pairs of individual snippets of voice and tried to come up with a confidence number for each whether it was Martin or Zimmerman.  How they came up with those numbers is pretty dicey, IMHO.  There also was very little in the report on whether it is valid at all to compare screams with normal speech.  Certainly a Frye hearing is needed to discuss that.  They themselves say (page 4)

Samples this brief rarely lead to attempts at speaker indentification.  Ordinarily, 10 words or 10 seconds of speech constitute a bare minimum.  However, they were the only "unknown" samples available and the task involved making a determination between but two speakers...

  I conclude from this they were under pressure to come up with identifications from individual snippet pairs, and against their better judgment complied.  And when is a conclusion that they don't have a match, just it is possible, ever allowed as evidence to convict beyond a reasonable doubt?

Based on the many analyses carried out, the undersigned has to conclude that, while there is evidence to suggest that Mr. Martin made the first two calls/cries (Nos one and eight) and that Mr. Zimmerman made those identified as 14 and 16, none of these conclusions reached the criterion for a match.

[nomatter_nevermind]

Hollien/Harnsberger disregards as unusable all the intelligible cries of 'help' or 'help me'. These are the six utterances they found usable:

wow, ow, cherp, wyra, owa, swa

[MJW]

In considering whether to allow Reich to testify, I wonder if a Frye hearing or sanity hearing is more appropriate.

[cboldt]

And when is a conclusion that they don't have a match, just it is possible, ever allowed as evidence to convict beyond a reasonable doubt?

No.  In order to be useful to the finder of fact, the expert has to reach a conclusion to at least a reasonable degree of certainty.  If the expert can't proffer that, then the expert isn't helpful.

[RickyJim]

From today's OS. 

But Zimmerman's lawyer says he may ask to delay the trial in light of the new analysis; The defense may need an additional expert to counter one of the new state reports, defense attorney Mark O'Mara told the Orlando Sentinel.
................
O'Mara called Reich's analysis flawed, and he may ask for a trial delay to find an expert to rebut him. The report by Hollien and Harnsberger, he said, was inconclusive. O'Mara said a defense expert has analyzed the same audio and concluded Zimmerman is the one calling for help.

One would think that the report from O'Mara's expert should appear on GZDocs soon.

[nomatter_nevermind]

Responding to post on another thread.

I think Zimmerman says that Martin was talking.  "Shut up," etc.

'Shut the f* up!' It's in the written statement (p. 3), the reenactment (9:15-25), and the CVSA (31:05-59).

In the CVSA, Zimmerman said that Martin said the same line at least twice, when Zimmerman first started yelling, and when Martin started to suffocate him. Only the first occasion is mentioned in the written statement, and only the second one in the reenactment.

IOW, that both of them were making sounds is not a point of contention.

Some have tried to impeach Zimmerman on this point, because the words he attributed to Martin aren't audible on the recording.

Whatever weight Hollien/Harnsberger has, it is corroborative of Zimmerman.

[nomatter_nevermind]

I've just started on Reich's report. He thinks the NEN call is a 911 call.

I know many people think this is trivial. I think it is significant evidence that the investigator isn't paying attention.