Author Topic: June 7th Hearing  (Read 13985 times)

0 Members and 1 Guest are viewing this topic.

Offline annoyedbeyond

  • Hero Member
  • *****
  • Posts: 1407
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #15 on: June 07, 2013, 08:20:35 AM »
How come no one is asking him why he owns the Easy Voice website? He's trying to say "they" as though EV is separate from him somehow.


Offline cboldt

  • Hero Member
  • *****
  • Posts: 1262
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #16 on: June 07, 2013, 08:34:30 AM »
West, on cross exam, elicits that Owen's math background is insufficient to explain how the algorithm of analysis works.  West offers Owen to explain what "Gaussian" means.  Owen fails at that.  West asks if the software has been submitted to NIST for evaluation.  Owen doesn't know if the package that he has has been submitted, but the software company has submitted to NIST.  Owen refers to a document from NIST to a person in the US who is in the software.  West asks if it can be provided to him.  Did Owen provide this to Mantei?  Yes, maybe Monday.  West also asks if Owen has the transcript - yes, it is embedded in the Soundforge output - as the sound plays, the "words" appear on the screen.  Mantei can;t reproduce this in court, lacking Soundforge.  West just wants to clarify that there is no written transcript, and asks Owen to repeat what he said.  Soundforge does not recognize speech - Owen typed in the words he thought he heard, as it played.  Owen identified 8 of 10 sounds as English words in the nature of "help" or "help me."

Objection from Mantei, I think, that this line of questioning is outside the scope of qualification, as this gets into word identification, and Owen will not be called to perform word identification.

Sound person on the stream has Nelson's microphone muted.  I can't make sense of what she's saying.  The issue is all around timing of testimony, and whether or not word identification is part of witness qualification under Frye.

West's argument is that he wants to probe whether or not word assignment (by the analyst) affects speaker ID, using the software.  West covers the historical ground as to the state's intention to use Owen at trial, his initial categorization was not as a testifying witness.

Offline cboldt

  • Hero Member
  • *****
  • Posts: 1262
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #17 on: June 07, 2013, 08:39:34 AM »
How come no one is asking him why he owns the Easy Voice website? He's trying to say "they" as though EV is separate from him somehow.

I think he's just a marketing agent for the software.  He testified that he does not know the math behind the analysis, and he demonstrated that he does not know the meaning of "Gaussian" (a math term) when West offered that as an easy question for Owen to demonstrate understanding of the analytical methods he was using.

Few minute recess as Owen e-mails material to the court.  The NIST review, as well as what passes for a transcript.

Mantei says that now what is being discussed is product (opinion), not methodology; and this hearing is supposed to be just about methodology.  I think West has the better argument, if word assignment has play in the methodology of speaker ID.

Offline RickyJim

  • Hero Member
  • *****
  • Posts: 1580
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #18 on: June 07, 2013, 08:41:42 AM »
I wonder what scientific work there is that proves there should be a particular correspondence between a person's ordinary speech, increased in frequency to scream level and the same person's true screams.

Offline annoyedbeyond

  • Hero Member
  • *****
  • Posts: 1407
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #19 on: June 07, 2013, 08:42:10 AM »
BTW cboldt, thanks for your notes. I have a ton of things I really need to get done that don't include watching this hearing (even though I haven't been able to tear myself away yet) and it's good to know there's a concise, even handed summary being posted.

Offline annoyedbeyond

  • Hero Member
  • *****
  • Posts: 1407
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #20 on: June 07, 2013, 08:44:46 AM »
I wonder what scientific work there is that proves there should be a particular correspondence between a person's ordinary speech, increased in frequency to scream level and the same person's true screams.

I don't guess there is any, or it would've been mentioned by now.


BTW RJ, you probably know this already but you can download a demo of the Easy Voice and watch it work.
I haven't bothered.

Offline RickyJim

  • Hero Member
  • *****
  • Posts: 1580
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #21 on: June 07, 2013, 08:53:04 AM »
Owen's conclusion that it was not Zimmerman's voice was obtained by noting absence of semitones in the scream recording, vs. Zimmerman's usual speaking voice.  He increased the pitch of the normal speaking voice to match the pitch in the screams.  With the basic pitch being equalized, the semitone peaks did not match, between the two samples.  Semitone is a musical term, that can be corrollated with "harmonics."  Generally, they are defined points between octaves of pitch.  With semitones, the octave is carved into 12 tones.  Harmonics naturally occur in vibrating objects, and the harmonics don;t necessarily land exactly on semitone frequencies.
Why should the semitone peaks match?  West should ask him for the science behind that assertion.

Offline annoyedbeyond

  • Hero Member
  • *****
  • Posts: 1407
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #22 on: June 07, 2013, 08:56:11 AM »
Why should the semitone peaks match?  West should ask him for the science behind that assertion.

Not saying it's science (fairly sure on my own research that it's not really--even though I'm less qualified than Owen is), but here are vids (produced by them, natch) explaining how things work.

http://www.easyvoicebiometrics.com/index.php?app=cms&ns=display&ref=videowalkthrough

Offline cboldt

  • Hero Member
  • *****
  • Posts: 1262
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #23 on: June 07, 2013, 08:59:01 AM »
Why should the semitone peaks match?  West should ask him for the science behind that assertion.

Yes, that's a good point.  He claimed to have studied the correlation in spectrum between normal speech and voice, but wasn't tasked with producing support for the contention that the semitone (frequency or pitch) peaks are the same between normal voice and screaming, in any given human.

Offline cboldt

  • Hero Member
  • *****
  • Posts: 1262
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #24 on: June 07, 2013, 09:01:22 AM »
BTW cboldt, thanks for your notes. I have a ton of things I really need to get done that don't include watching this hearing (even though I haven't been able to tear myself away yet) and it's good to know there's a concise, even handed summary being posted.

Thanks.  I miss quite a bit, and occasionally mistake events 180 degrees from what really happened.  I have a couple kids graduating from high school today, and I'm sure I'll be fading out of sight here before Nelson's Court Festivities conclude.

Offline jjr495

  • Full Member
  • ***
  • Posts: 118
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #25 on: June 07, 2013, 09:21:28 AM »
Owen said the program would not try to make an id on less than 16 seconds. So he doubled his eight second compilation. I would think there is a reason that the software creators had a reason for the 16 second limit. Why not take 1 second and repeat it 16 times.

Offline RickyJim

  • Hero Member
  • *****
  • Posts: 1580
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #26 on: June 07, 2013, 09:53:41 AM »
Owen just deFryed himself.  He admits he is the only one to use this method to identify screams.

Offline RickyJim

  • Hero Member
  • *****
  • Posts: 1580
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #27 on: June 07, 2013, 09:58:17 AM »
How the heck can Nelson accept this guy as passing Frye?

Offline cboldt

  • Hero Member
  • *****
  • Posts: 1262
  • Rate Post +0/-0
Re: June 7th Hearing
« Reply #28 on: June 07, 2013, 10:11:09 AM »
West hammers Owen on looping in order to get a long enough sample for the software to process.  He is looking for authority that justifies looping, and notes the ramifications to conclusions caused by looping.  Owen has never done this "looping" before.  Another ding in the Frye framework - looping is a new and novel approach to use of the biometric software.  Owen has no authority that he is able to produce that explains the ramifications of looping against the reliability of the conclusion produced by the analytical method.

I don't think Owen knows why the 16 seconds of sample is at least preferred, if not required.  Is the sample intended to capture a range of vowel/consonant sounds?  West asks what is the minimum amount of speech required to support a reliable analysis?  Owen says 20 words.  West asks if 20 words includes the same word over and over, or is there a need for different words?  Owen says the machine is comparing 4 seconds of screams with 6 seconds of Zimmerman reenactment.  West gets Owen to admit that he (Owen) doesn't know which pieces of the (scream) loop are selected by the software.

Offline Cylinder

  • Hero Member
  • *****
  • Posts: 730
  • Rate Post +0/-0
  • IANAL
Re: June 7th Hearing
« Reply #29 on: June 07, 2013, 10:16:36 AM »
This goes directly to Ramirez. In that case, a tool mark match was a match because the expert said it was - there was no objective standard. Owen is saying basically the same thing.

 

Site Meter
click
tracking