Author Topic: June 8th Hearing  (Read 12506 times)

0 Members and 1 Guest are viewing this topic.

Offline cboldt

  • Hero Member
  • *****
  • Posts: 1262
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #15 on: June 08, 2013, 08:21:13 AM »
I'm kind of looking forward to cross. I'm curious how Mantei approaches him.

Cookie cutter says, he'll asks only about the methods, and not application of the method to the evidence in this case, and he'll ask if different experts can reasonably disagree given the same evidence.

Offline annoyedbeyond

  • Hero Member
  • *****
  • Posts: 1407
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #16 on: June 08, 2013, 08:21:26 AM »
looping and raising pitch aren't going to be accepted in the scientific community.

Offline nomatter_nevermind

  • Hero Member
  • *****
  • Posts: 5449
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #17 on: June 08, 2013, 08:28:13 AM »
Critiquing Reich, explaining why the very idea of a voice print is a false analogy.

He's a perfect expert. He knows his stuff, and he can make it clear to layfolk.

Offline nomatter_nevermind

  • Hero Member
  • *****
  • Posts: 5449
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #18 on: June 08, 2013, 08:30:45 AM »
No accepted methodology for determining age from voice. A 50-year-old can scream like a child.

Offline RickyJim

  • Hero Member
  • *****
  • Posts: 1580
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #19 on: June 08, 2013, 08:46:08 AM »
Mantai tried to get French to say that the screams couldn't come from somebody who was being smothered.  No dice.

Offline annoyedbeyond

  • Hero Member
  • *****
  • Posts: 1407
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #20 on: June 08, 2013, 08:54:10 AM »
Hilarious that Mantei is still trying to paint Traytray as a child.

Especially in light of ABC releasing his voice samples yesterday.


Offline cboldt

  • Hero Member
  • *****
  • Posts: 1262
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #21 on: June 08, 2013, 08:56:55 AM »
French is not entirely sure what aural spectrograph is.  If if is like the discredited voiceprint analysis, combined with listening, it is not reliable.  Never been used in the UK.  No speech scientist anywhere in the world would condone it.  West asks if there is any such thing as a voiceprint.  Dr. French says "no."  There is never an exact match between individual signatures, nor between individual words being said.  Each speech event is unique.  I think the reference is just to compare the time/level signals - literally measuring the sound wave/pressure over time.

Have you reviewed Dr. Reich's report?  Yes.  It bears similarity to acoustic listening method French already described, farther from the voiceprint analysis.  He finds a number of things in Reich's report disturbing - for example, finding certain words (stop) via resonant frequencies relative to the normal male.  Reich suggests that because the value (pitch) is 10% higher, that the vocal tract must be that of Martin, due to him not being full grown.  French says there is not a correlation between growth (overall maturity), and vocal tract maturity, and speakers age.  There is NO accepted data to support finding age.  West asks if a 28 year old could make that sound.  French says a 50 year old could.  Goes on to detail the physics, moth would be more open.  West asks if French has any experience in phonetics.  Yes, French's postgrad was in phonetics, and he has had ongoing training in this area, and is called on to train other people in phonetic analysis.

Back to Reich hearing the word "stop."  Part two, was this related to screams in particular, or something else?  French says yes, there are remarks about words head, "these shall be," etc.  Not a single one of those does Dr. French hear.  He hears breathing or other random sounds, but he can't accept that any of those sounds is speech.  West asks if there is a recognized phenomenon where listeners under high amplification.  Dr. Frazer in Australia has a body of work where people mistake for speech, non-speech sounds or patterns.  Is that phenomenon recognized in your work?  yes, we are aware of this and endeavor to eliminate.  How do you deal with this?  Dr. French says other than overall caution, there is no way.  He says in this recording, if you didn't know from outside information, you wouldn't know that the person was screaming in English, was a male or female.

BTW, Mantei objected to the line of questioning about predisposition or inclination to hear words or speech where none is being uttered, on the grounds that it aimed to obtain a psychiatric impression of the listener.  Objection overruled, and Dr. French said he would "be careful to not trespass on the grounds your opponent has identified."

West wants a few moments to make sure he's covered the bases.  Asks maybe for a recess.  Nelson says no recess until we are done with this witness.

Mantei up.

Your first testimony in a US court?  Yes.  You do not hear "you're going to die tonight motherfscker" or "you got me"?  No, I don't hear those things.  French isn't saying that there is nothing that might not be construed as that (with imagination), but he doesn't hear it.  Mantei goes on to describe the recording arrangement - 40 feet from outdoors to indoors, phone, recorded, competition with other voices on the tape.  How much effort or power would be required to generate the signal from the outside.  French says not enough info - depends on head, mouth placement of screamer; microphone on phone direction, how close the phone is to the speaker's face, quality of phone line.  Can't give a meaningful view of that.

How about whether or not a scream was smothered or muffled.  French could not tell either way.

Is voice recognition a generally valid undertaking?  Yes, we evaluate a great number of samples.  Mantei says you do not have a minimum amount of words or time.  Yes, true.  Mantei asks if French listened to the sounds without assistance, and also with SoundForge and other analytical tools.  French says yes, he did.  SoundForge is described as a blunt tool.  Mantei says you measured the pitch? Found it to be 300-900 Hz?  French says he may have recalculated these values with assistance from his PhD student and the technique she uses.  Average pitch 526 Hz, Max 843 Hz, minimum 342 Hz.  Mantei asks why did you do that?  French explains, his first cut was use of default program, and the PhD student has techniques that ameliorate errors that the speech software tends to make in spectral analysis.  French makes a great point that the difference would not change his conclusion.  Just because you change something to improve your result isn't something that should concern scientists, should it?

Mantei fairly correctly summarizes French, that none of the processes will deliver reliable results; French adds that biometric analysis is completely inappropriate.  Mantei tries to say that none of the methods are bad; but the issue is the sample.  French says that the methodology of Owen would not be accepted by the scientific community (biometric).

Mantei goes on to the vocal tract development.  French says that this comparison is invalid for screaming.  Mantei asks if adolescents and men have deeper voices as they age.  Mantei is done.

Redirect by West.

Would French describe the process of voice change in men.  Typically in early teen years, he is not an expert on this, but typically between 12-16, maybe a little older.  Lowering of larynx, enlargement of sinus cavities, enlargement of larynx.  Is this time of change associated with other physiological changes?  Yes, with puberty.  West says what about facial hair, pubic hair - French says this strays outside of his area of expertise, and a juror or judge could answer those questions as well as French could.  West asks if French has any data / conclusion as to whether or not Martin had undergone puberty.  Based on the video of martin, French thinks Martin is in or past puberty.  Witness is over.

15 minute recess.

Mantei's cross exam was lame.

Offline nomatter_nevermind

  • Hero Member
  • *****
  • Posts: 5449
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #22 on: June 08, 2013, 09:01:59 AM »
Your first testimony in a US court?  Yes.

I didn't hear the end of that question. I thought Mantei had asked if it was French's first testimony in a Florida court, because I thought at the beginning French said he had testified twice in the U.S. Maybe he was talking about some other kind of professional work?

Offline RickyJim

  • Hero Member
  • *****
  • Posts: 1580
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #23 on: June 08, 2013, 09:02:37 AM »
Cbolt, are you a trained court reporter?  I certainly can't do what you can do.  Very impressive!  By the way, I am surprised that Mantei didn't ask French how much he was being paid. 

Offline Philly

  • Jr. Member
  • **
  • Posts: 37
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #24 on: June 08, 2013, 09:11:23 AM »
French acknowledged under cross that he couldn't make out the words "You're going to die tonight" or "Shut up!" - phrases that Zimmerman has asserted that Trayvon said during the struggle.

I was surprised West didn't revisit this when he reengaged the expert.

The expert had noted that the direction one is facing could make a big difference when answering the "how much power would it take for the screams to be audible" questions.  I'd also be curious to learn more about how phone compression algorithms might end up selectively filtering sounds, and whether this could explain why a scream would be more audible than normal speech/threats.

Offline nomatter_nevermind

  • Hero Member
  • *****
  • Posts: 5449
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #25 on: June 08, 2013, 09:14:14 AM »
French acknowledged under cross that he couldn't make out the words "You're going to die tonight" or "Shut up!" - phrases that Zimmerman has asserted that Trayvon said during the struggle.

I was surprised West didn't revisit this when he reengaged the expert.

It's a red herring.

Its purpose may have been to distract West into wasting time on it.

Offline RickyJim

  • Hero Member
  • *****
  • Posts: 1580
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #26 on: June 08, 2013, 09:18:59 AM »
I think it is sort of obvious that the amplitude is a very major factor, besides the ones French mentioned, in what a phone is the house might pick up.  Screams have a much higher energy than the other things said that night.  I wonder if the prosecution has people in the wings who also hear what Reich claimed to hear.  The only thing I am really getting out of all this is that the prosecution's case is really as bad as we all thought.

Offline cboldt

  • Hero Member
  • *****
  • Posts: 1262
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #27 on: June 08, 2013, 09:19:17 AM »
Cbolt, are you a trained court reporter?  I certainly can't do what you can do.  Very impressive!  By the way, I am surprised that Mantei didn't ask French how much he was being paid.

No, just taking notes on the fly.  I type while I listen, almost always paraphrasing (hopefully, accurately enough).  Many years in technical work, trying to elicit information that is useful and factual, and then having to report or transmit that in many directions.

Thank you for the kind compliment.  I do this for fun, so when I get bored with the case, I disappear ;-)

Offline nomatter_nevermind

  • Hero Member
  • *****
  • Posts: 5449
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #28 on: June 08, 2013, 09:21:53 AM »

G. R. Doddington.

Jeralyn has posted short profiles of the three defense experts.

Offline cboldt

  • Hero Member
  • *****
  • Posts: 1262
  • Rate Post +0/-0
Re: June 8th Hearing
« Reply #29 on: June 08, 2013, 09:37:42 AM »
Doddington.  Electrical engineer, Texas Instruments, has done work with NIST and others.  Undergrad at U Florida, PhD from U Wisc, Bell Labs for PhD, dissertation was on speaker recognition.  The title was "Practical method of speaker verification" or similar.  Work with TI was speech and speaker recognition, much under contract with Air Force.  The witness doesn't know why the AF and TI had an interest in speech and speaker recognition.  TI is a bunch of cowboys down there, and the management had reasons only they knew (witness named TI president, I forgot the name).

From a security standpoint, speaker trying to gain access to a resource, and then NSA, trying to track a person who does not want to be found.  So, diametrically opposed parameters.  Doddington's work was to the "gaining access" parameter, where the software aimed to match the speaker, with high reliability.  Entry control to enable a door to open, for example.  An operational system of this nature was installed at a TI computer center.

Doddington is obsessed (his word) with calibrating the performance of technology.  He got into "what is the performance of speech recognizers," which was a cover article in a 1981 IEEE "Spectrum" article.  This was the first generation of speech recognizers that was being fielded.  The performance being claimed was favorable to the system, but by very large factors, and this bothered the witness, because they provided far poorer performance.  In a nutshell, he says, speech is not like fingerprints, or iris scans, or DNA, which are physical artifacts.  Speech is a performance, like a dance.  It is an energy signal.  Speech is a learned performance, and individuals have idiosynchrocies, but any individual is variable in his own "performances" of speech.  Doddington was in charge of TI's speech scientist before he left - he ran the speech research lab, small but very good.

Went to SRI international in Menlo Park, CA.  Stanford Research Institute (no longer with Stanford).  Joined as senior scientist, was borrowed by federal government to work with DARPA.  The focus of that work was program manager to all of human language technology.

I'm away for awhile .... this is the defense pretty much just piling on.  State's experts aren't ;-)


 

Site Meter
click
tracking