i'd be interested to hear opinions here (professional and otherwise) on the reliability of determining who was screaming. given the recording was through a cellular network and passed through codecs, i'd say it's damn near impossible. even if you had a perfectly clear recording, how could you classify a scream? it seems the parameters of speech would change drastically when the subject is under such duress. -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
the zimmerman case and forensic audio analysis
Started by ●June 8, 2013
Reply by ●June 8, 20132013-06-08
On Fri, 07 Jun 2013 23:26:39 -0400, Randy Yates <yates@digitalsignallabs.com> wrote:>i'd be interested to hear opinions here (professional and otherwise) on >the reliability of determining who was screaming. > >given the recording was through a cellular network and passed through >codecs, i'd say it's damn near impossible. even if you had a perfectly >clear recording, how could you classify a scream? it seems the >parameters of speech would change drastically when the subject is under >such duress.Hi Randy, I remember, from years ago, the notion that something called "voice prints" were almost as reliable as fingerprints in identifying someone. Now I don't know diddly about voice prints, but I'll bet we hear something about voice prints in the next month with regard to the Zimmerman trial. By the way Randy, I predict that Zimmerman is going to be sentenced to prison. It's my guess that, regardless of any photographs, testimony, or any evidense whatsoever, he's on his way to the Grey-bar Hotel. That's just my prediction. [-Rick-]
Reply by ●June 8, 20132013-06-08
Rick Lyons <R.Lyons@_BOGUS_ieee.org> writes:> On Fri, 07 Jun 2013 23:26:39 -0400, Randy Yates > <yates@digitalsignallabs.com> wrote: > >>i'd be interested to hear opinions here (professional and otherwise) on >>the reliability of determining who was screaming. >> >>given the recording was through a cellular network and passed through >>codecs, i'd say it's damn near impossible. even if you had a perfectly >>clear recording, how could you classify a scream? it seems the >>parameters of speech would change drastically when the subject is under >>such duress. > > Hi Randy, > I remember, from years ago, the notion that > something called "voice prints" were almost as > reliable as fingerprints in identifying someone. > Now I don't know diddly about voice prints, but > I'll bet we hear something about voice prints > in the next month with regard to the Zimmerman trial.OK, that's a good search term - thanks!> By the way Randy, I predict that Zimmerman is going > to be sentenced to prison. It's my guess that, > regardless of any photographs, testimony, or any > evidense whatsoever, he's on his way to the > Grey-bar Hotel. That's just my prediction."Don't bother me with the facts, my mind is made up!"? -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●June 8, 20132013-06-08
Rick Lyons <R.Lyons@_BOGUS_ieee.org> writes:> On Fri, 07 Jun 2013 23:26:39 -0400, Randy Yates > <yates@digitalsignallabs.com> wrote: > >>i'd be interested to hear opinions here (professional and otherwise) on >>the reliability of determining who was screaming. >> >>given the recording was through a cellular network and passed through >>codecs, i'd say it's damn near impossible. even if you had a perfectly >>clear recording, how could you classify a scream? it seems the >>parameters of speech would change drastically when the subject is under >>such duress. > > Hi Randy, > I remember, from years ago, the notion that > something called "voice prints" were almost as > reliable as fingerprints in identifying someone. > Now I don't know diddly about voice prints, but > I'll bet we hear something about voice prints > in the next month with regard to the Zimmerman trial.I guess part of what's bothering me is that a couple of these guys are claiming with great confidence that it's not Zimmerman's voice. Are these truly claims that conscientious scientists/engineers would make? Are they THAT sure of their analysis? -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●June 8, 20132013-06-08
On Sat, 08 Jun 2013 12:41:45 -0400, Randy Yates <yates@digitalsignallabs.com> wrote:>Rick Lyons <R.Lyons@_BOGUS_ieee.org> writes: > >> On Fri, 07 Jun 2013 23:26:39 -0400, Randy Yates >> <yates@digitalsignallabs.com> wrote: >> >>>i'd be interested to hear opinions here (professional and otherwise) on >>>the reliability of determining who was screaming. >>> >>>given the recording was through a cellular network and passed through >>>codecs, i'd say it's damn near impossible. even if you had a perfectly >>>clear recording, how could you classify a scream? it seems the >>>parameters of speech would change drastically when the subject is under >>>such duress. >> >> Hi Randy, >> I remember, from years ago, the notion that >> something called "voice prints" were almost as >> reliable as fingerprints in identifying someone. >> Now I don't know diddly about voice prints, but >> I'll bet we hear something about voice prints >> in the next month with regard to the Zimmerman trial. > >I guess part of what's bothering me is that a couple of these guys are >claiming with great confidence that it's not Zimmerman's voice. Are >these truly claims that conscientious scientists/engineers would make? >Are they THAT sure of their analysis?Hi Randy, Again, I predict that Zimmerman is going to be sentenced to prison. It's my guess that, regardless of any photographs, testimony, or any evidense whatsoever, he's on his way to the Grey-bar Hotel. That's just my prediction. See Ya', [-Rick-]
Reply by ●June 8, 20132013-06-08
On Sunday, June 9, 2013 12:40:24 PM UTC+12, Rick Lyons wrote:> On Sat, 08 Jun 2013 12:41:45 -0400, Randy Yates > > <yates@digitalsignallabs.com> wrote: > > > > >Rick Lyons <R.Lyons@_BOGUS_ieee.org> writes: > > > > > >> On Fri, 07 Jun 2013 23:26:39 -0400, Randy Yates > > >> <yates@digitalsignallabs.com> wrote: > > >> > > >>>i'd be interested to hear opinions here (professional and otherwise) on > > >>>the reliability of determining who was screaming. > > >>> > > >>>given the recording was through a cellular network and passed through > > >>>codecs, i'd say it's damn near impossible. even if you had a perfectly > > >>>clear recording, how could you classify a scream? it seems the > > >>>parameters of speech would change drastically when the subject is under > > >>>such duress. > > >> > > >> Hi Randy, > > >> I remember, from years ago, the notion that > > >> something called "voice prints" were almost as > > >> reliable as fingerprints in identifying someone. > > >> Now I don't know diddly about voice prints, but > > >> I'll bet we hear something about voice prints > > >> in the next month with regard to the Zimmerman trial. > > > > > >I guess part of what's bothering me is that a couple of these guys are > > >claiming with great confidence that it's not Zimmerman's voice. Are > > >these truly claims that conscientious scientists/engineers would make? > > >Are they THAT sure of their analysis? > > > > Hi Randy, > > Again, I predict that Zimmerman is going > > to be sentenced to prison. It's my guess that, > > regardless of any photographs, testimony, or any > > evidense whatsoever, he's on his way to the > > Grey-bar Hotel. That's just my prediction. > > > > See Ya', > > [-Rick-]seems it was self defense http://en.wikipedia.org/wiki/Shooting_of_Trayvon_Martin
Reply by ●June 8, 20132013-06-08
Rick Lyons <R.Lyons@_BOGUS_ieee.org> writes:> On Sat, 08 Jun 2013 12:41:45 -0400, Randy Yates > <yates@digitalsignallabs.com> wrote: > >>Rick Lyons <R.Lyons@_BOGUS_ieee.org> writes: >> >>> On Fri, 07 Jun 2013 23:26:39 -0400, Randy Yates >>> <yates@digitalsignallabs.com> wrote: >>> >>>>i'd be interested to hear opinions here (professional and otherwise) on >>>>the reliability of determining who was screaming. >>>> >>>>given the recording was through a cellular network and passed through >>>>codecs, i'd say it's damn near impossible. even if you had a perfectly >>>>clear recording, how could you classify a scream? it seems the >>>>parameters of speech would change drastically when the subject is under >>>>such duress. >>> >>> Hi Randy, >>> I remember, from years ago, the notion that >>> something called "voice prints" were almost as >>> reliable as fingerprints in identifying someone. >>> Now I don't know diddly about voice prints, but >>> I'll bet we hear something about voice prints >>> in the next month with regard to the Zimmerman trial. >> >>I guess part of what's bothering me is that a couple of these guys are >>claiming with great confidence that it's not Zimmerman's voice. Are >>these truly claims that conscientious scientists/engineers would make? >>Are they THAT sure of their analysis? > > Hi Randy, > Again, I predict that Zimmerman is going > to be sentenced to prison. It's my guess that, > regardless of any photographs, testimony, or any > evidense whatsoever, he's on his way to the > Grey-bar Hotel. That's just my prediction.Hi Rick, I'm afraid you may be right. But I still hold people in our field (engineering in general and a special area of signal processing) accountable for their statements and their integrity. Now I am NOT a "forensic audio expert," but I do have some experience in audio engineering, and from what I know, there is no way in hell someone could claim with confidence the authentication of the voice in this scenario. I suspect agendas are at play, or outright corruption. -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●June 9, 20132013-06-09
On 6/8/2013 10:48 PM, Randy Yates wrote:> Rick Lyons <R.Lyons@_BOGUS_ieee.org> writes: > >> On Sat, 08 Jun 2013 12:41:45 -0400, Randy Yates >> <yates@digitalsignallabs.com> wrote: >> >>> Rick Lyons <R.Lyons@_BOGUS_ieee.org> writes: >>> >>>> On Fri, 07 Jun 2013 23:26:39 -0400, Randy Yates >>>> <yates@digitalsignallabs.com> wrote: >>>> >>>>> i'd be interested to hear opinions here (professional and otherwise) on >>>>> the reliability of determining who was screaming. >>>>> >>>>> given the recording was through a cellular network and passed through >>>>> codecs, i'd say it's damn near impossible. even if you had a perfectly >>>>> clear recording, how could you classify a scream? it seems the >>>>> parameters of speech would change drastically when the subject is under >>>>> such duress. >>>> >>>> Hi Randy, >>>> I remember, from years ago, the notion that >>>> something called "voice prints" were almost as >>>> reliable as fingerprints in identifying someone. >>>> Now I don't know diddly about voice prints, but >>>> I'll bet we hear something about voice prints >>>> in the next month with regard to the Zimmerman trial. >>> >>> I guess part of what's bothering me is that a couple of these guys are >>> claiming with great confidence that it's not Zimmerman's voice. Are >>> these truly claims that conscientious scientists/engineers would make? >>> Are they THAT sure of their analysis? >> >> Hi Randy, >> Again, I predict that Zimmerman is going >> to be sentenced to prison. It's my guess that, >> regardless of any photographs, testimony, or any >> evidense whatsoever, he's on his way to the >> Grey-bar Hotel. That's just my prediction. > > Hi Rick, > > I'm afraid you may be right. But I still hold people in our field > (engineering in general and a special area of signal processing) > accountable for their statements and their integrity. > > Now I am NOT a "forensic audio expert," but I do have some experience in > audio engineering, and from what I know, there is no way in hell someone > could claim with confidence the authentication of the voice in this > scenario. > > I suspect agendas are at play, or outright corruption.I understand voiceprints using high-quality signals. When the SNR gets low, the results, to analogize from fingerprints, are "smudged". Is there any forensic value at all to signals that have been squeezed through mobile phones? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●June 9, 20132013-06-09
Jerry Avins <jya@ieee.org> wrote:> On 6/8/2013 10:48 PM, Randy Yates wrote:(snip on Zimmerman)>> I'm afraid you may be right. But I still hold people in our field >> (engineering in general and a special area of signal processing) >> accountable for their statements and their integrity.>> Now I am NOT a "forensic audio expert," but I do have some experience in >> audio engineering, and from what I know, there is no way in hell someone >> could claim with confidence the authentication of the voice in this >> scenario.>> I suspect agendas are at play, or outright corruption.> I understand voiceprints using high-quality signals. When the SNR gets > low, the results, to analogize from fingerprints, are "smudged". Is > there any forensic value at all to signals that have been squeezed > through mobile phones?Most of the important parts of voice aren't all that high frequency, which is why the telephone system can get away with 4kHz. Seems to me that it would be easier to show that two signals were from differnet people than to show that they were the same person, but still I will leave it to experts. If someone has looked at many different voiceprints, screaming or not, cellphone or not, they should have good statistics on the analysis that we don't have. Statistics will be more important than DSP, it seems to me. -- glen
Reply by ●June 9, 20132013-06-09
On Monday, June 10, 2013 5:31:05 AM UTC+12, glen herrmannsfeldt wrote:> Jerry Avins <jya@ieee.org> wrote: > > > On 6/8/2013 10:48 PM, Randy Yates wrote: > > > > (snip on Zimmerman) > > > > >> I'm afraid you may be right. But I still hold people in our field > > >> (engineering in general and a special area of signal processing) > > >> accountable for their statements and their integrity. > > > > >> Now I am NOT a "forensic audio expert," but I do have some experience in > > >> audio engineering, and from what I know, there is no way in hell someone > > >> could claim with confidence the authentication of the voice in this > > >> scenario. > > > > >> I suspect agendas are at play, or outright corruption. > > > > > I understand voiceprints using high-quality signals. When the SNR gets > > > low, the results, to analogize from fingerprints, are "smudged". Is > > > there any forensic value at all to signals that have been squeezed > > > through mobile phones? > > > > Most of the important parts of voice aren't all that high frequency, > > which is why the telephone system can get away with 4kHz. > > > > Seems to me that it would be easier to show that two signals were > > from differnet people than to show that they were the same person, > > but still I will leave it to experts. > > > > If someone has looked at many different voiceprints, screaming or > > not, cellphone or not, they should have good statistics on the > > analysis that we don't have. Statistics will be more important > > than DSP, it seems to me. > > > > -- glenhttp://www2.lti.cs.cmu.edu/Research/Thesis/QinJin.pdf
