I think the defense should have shown Figures 3, 5, and 6 from this article.
Fig 3 shows how pitch dramatically changes under different types of speech. Fig 5 shows how most but not all of the spectral features change. Fig 6 show how the first formant changes, but the others not as much.
So when Owen adjusted the frequency of the exemplar to match the pitch of the screams, he completely screwed up all of the other spectral signatures that his biometric system measures and compares.
So it was preordained that the exemplar and screams would not match. That is much worse than a 50% random coin flip. For George it was heads you lose tails the state wins because of Owens idiotic methodology.
Probably too late for this, but given everything Owen did to the samples of George's voice to get his results, I'm wondering if the defense could find someone with the same software he's peddling and have them pitch up a sample of Trayvon's voice, then compare the results to the results of an analysis of the same sample unaltered.
I suspect they wouldn't get any better match than what Owen did with George's voice, and it's one of those in-court demonstrations that makes the prosecution's case look bad.
And did Owen pitch up five
semitones? I think that's from C to F on a piano keyboard.