Report says it wasn’t Zimmerman yelling on the 911 call
(Note: Alexis Zarkov has taken a look
at the voice recognition software that was used here, and finds a lot of room for uncertainty about its touted “certain” conclusion that the man yelling on the tape is not Zimmerman.)
Jeff C. writes:
The weight shifts.
I read this article yesterday, and I would be cautious about signing on to its stated conclusions, for these reasons:
First, it is a matter of percentage of match—something relative, not something absolute. Yet they are presenting it in assured terms as though it were absolute. They said it was a 48 percent match, and should be much higher. That is something that can be easily jiggered.
Second, they don’t have a recording of Martin. This is based only on Zimmerman’s voice.
Third and most important, the liberal media lie about EVERYTHING. After NBC News’ doctoring of the audio tape of Zimmerman’s 911 call, which had the whole world believing that Zimmerman said something that he didn’t say, NOTHING the liberal media say that is incriminatory of Zimmerman should be accepted on face value. It has to be clearly established as true, not just something claimed in an article.
Finally, as I’ve said from the start, we do not know what happened. We have no intelligible step by step account of the movements and actions of the two principals. Is it possible that Zimmerman provoked the confrontation? Yes. Is it possible that Zimmerman unnecessarily and without justification used fatal force? Yes. Is it even possible that Zimmerman was not in danger from Trayvon at the moment he pulled the trigger and that he murdered him? Yes. Is it possible that the police were too quick to release Zimmerman? Yes. All those things are possible, though, in my opinion, based on the facts that we do have, they are much less likely than their respective opposites; and, with regard to the possibility of murder, far less likely. We don’t have a complete picture of what happened that night. But we do know certain things. We know that there has been a national lynch mob against George Zimmerman (and against the racist white America that he supposedly represents), and a national beatification of Trayvon Martin (and of the innocent victimized black America that he supposedly represents). We know that from the start the media and the politicians have built up a lying picture of this event in which basic information tending to exculpate Zimmerman has been covered up and a falsely innocent picture of Martin has been promoted. These are facts, and it is on these facts that most of my commentary has been focused.
Alexis Zarkov writes:
This is bad news for Zimmerman as it seemingly contradicts his story. If he’s arrested then his lawyer needs to get his own forensic expert. However, as Mr. Auster points out, we can’t trust the liberal media, so I decided to do a little probing on my own.
According to the Orlando Sentinel article, Tom Owen of Owen Forensic Services used a forensic software product called Easy Voice Biometrics to conclude that the person screaming on the recording was not Zimmerman. Can we trust Mr. Owen? Does he have good credentials? I went to the Owen Forensic Services website, and indeed he has good credentials within the universe of forensic experts. His conclusion would carry weight in a court proceeding. However reading his curriculum vitae I see he has only a BA in history, and his experience indicates he’s more of an audio technician than an engineer. Essentially he’s a user of forensic products, and I suspect he lacks an in-depth understanding of how these products work.
Next I went to the Easy Voice Biometrics website. They sell a software product for about $5,000 that tests one or more voice recordings against a recording of an unknown speaker, and reports the amount of match as a percentage. But the website doesn’t (unless I missed it) tell us the theory and methodology behind the software. If the company said, “That’s proprietary,” then I would not take them seriously. Moreover they don’t tell us how to turn the match percentage into a probability.
Note also that Owen is reporting the lack of a match, not a match. I suspect this software functions as a statistical hypothesis testing device. It sets some match threshold, and if a match exceeds the threshold, it declares a positive (the unknown and known speakers are the same). If the match falls below, then it reports a negative. So to understand how well Easy Voice Biometrics works in general, we need two numbers: (1) the probability of a false positive, (2) the probability of a false negative. For Zimmerman we need to know (2). Had there been a match we would need to know (1). I would also like to know how well the software does when comparing a normal voice to a scream. It’s possible that a scream is so different, that you would virtually always get a low match percentage. But I don’t know without information from the company. It’s possible for a technique to do well on (1) but poorly on (2).
If I could get the recording of the scream, then I might be able to do my own analysis.
Thanks for doing this research. Of course we don’t have the answer, but your examination shows how much room there is in this claim for mistakes, false guesses, or even just plain b.s.
It’s remarkable how, no matter how many times various touted “scientific” findings have been exposed as self-interested or ideologically motivated bloviating (remember the 47 million year old primate “Ida,” hyped as the “missing link” and the “human-like” “human ancestor” that “revolutionized” our understanding of evolution, and whose “human-like” features turned out to be the features that all primates have in common, or the “first” European who just happened to be a Negro?), people still automatically believe the next “scientific” finding they read about in the press. It’s as though people living under Communism did not understand (or kept forgetting) that they lived under an ideological state, and believed that everything the regime told them was the truth. Well, we too live under an ideological state, which means not just the government, but all the institutions supporting the governing ideology. It is time for us to internalize what this means.
Neil Parille writes:
The Orlando Sentinel story about the voice doesn’t impress me. I don’t get the impression that this is an exact science like fingerprints or DNA.
“I believe that’s Trayvon Martin in the background, without a doubt,” Primeau says, stressing that the tone of the voice is a giveaway. “That’s a young man screaming.” Funny, I was listening to an interview with a neighbor of Zimmerman’s today who said he was a good guy. Until the interviewer mentioned the interviewee’s”wife” I wasn’t sure if the interviewee was a he or she.
From what I have read, Zimmerman said all along that he was screaming for help. There is an eyewitness who supports this. Zimmerman’s statement came before anyone knew that there was a a 911 call which recorded the incident. If Zimmerman engaged in premeditated murder as the left suggests, why did he call the police? You don’t have the police on the line when you are planning on killing someone. [LA replies: That’s not necessarily true. There was a case, covered on Snapped and other TV crime shows, in which a husband murdered his wife by pushing the car she was in over a cliff, and tried to establish his innocence by making a 911 call in which he pretended to have been thrown from the car and be hanging on the cliff in fear of his life. He put on this big dramatic act on the 911 call which the police eventually saw through.]
From everything I’ve read, the facts are more consistent with Martin being aggressor. On the other hand, this case is like trying to figure out a ten chapter book knowing only the first and last chapters.
Evan H. writes:
Regardless of what Tom Owen knows about audio forensics, he doesn’t seem to understand basic statistical inference. This is an example of how easy it is to mislead the layman with statistical arguments. Given that a 90 percent match is supposedly needed to prove that a voice sample matches an individual, and that only a 48 percent match was found for Zimmerman, all we can say is “the software was unable to prove that Zimmerman’s voice is on the recording.” Note that this statement is profoundly different from “the software proved that Zimmerman’s voice was not on the recording.” This is a very basic Stats 101 error that is made all the time by people with no statistical training, so I’m very surprised to hear it coming from a so-called expert. Furthermore, if we don’t know the rate of false negatives, we actually don’t even know how likely it is that Zimmerman’s voice actually is on the recording.
Consider this—suppose Owen runs the test using Martin’s voice and finds a 20 percent match, or a 48 percent match, or a 70 percent match (anything below the 90 percent threshold). Would we then conclude with “scientific certainty” that neither man’s voice is on the recording?
all we can say is “the software was unable to prove that Zimmerman’s voice is on the recording.” Note that this statement is profoundly different from “the software proved that Zimmerman’s voice was not on the recording.”
Thank you for this. I knew as I was reading the article that something wasn’t right about it, and I think you’ve put your finger on it.
Jim H. in Oregon writes:
I came across an interesting discussion about one of the fellows who did a voice analysis. Here are a few snippets:
Now, let’s get to Mr. Owen. Is he really a “forensics expert?”
Anyway, you can read the whole discussion here.
Well, no. He has a B.A. in History, not forensics.
Well, surely since he’s an expert, he went to Harvard or Yale or Oxford, right?
Well, no he went to Bellarmine College in Kentucky. You’ve heard of Bellarmine College, right?
Well, but surely he’s worked for the FBI or CIA as a forensics examiner?
No, but he did work for the New York Public Library in charge of their Rogers and Hammerstein archive—coughgaycough.
But surely he’s written some books about forensics that the FBI and CIA use, right?
No, but he has written extensively about the banjo and did a coloring book….
But surely they have an advanced forensic acoustics labratory where they conduct their forensic analysis right?
No, they just have a computer in the basement of their house in New Jersey right next to the railroad tracks.
But surely he’s a member of the American Board of Recorded Evidence.
Yes, he and his wife run the outfit. He’s the chairman, and she’s a board member. The Board is run out of a weight-loss clinic in Springfield, Missouri. They certified themselves experts by creating this fake board.
They do tend to testify in a lot of cases if you pay their fees and travel expenses, though.
Now, let’s examine the software they used to perform their analysis. How good is it?
Well, given two KNOWN samples of speech by President Richard M. Nixon, the software only matched two recordings to a 86% match.
Remember, Mr. Owen said he’d expect a 90% match before he concluded that the voices were identical. But in the software makers own demonstration of the product, using two KNOWN samples of speech from the same famous person, there was only an 86% match.
They lie about the nature of their testing facilities
Owen Forensic Services, L.L.C., maintains a state of the art facility for the purposes of digital audio enhancement, digital video enhancement, digital audio and video authenticity analysis, voice identification and media/data recovery.
Bull***t. They have a home in New Jersey and conduct their “analysis” from a desktop computer in their basement.
If it turns out that the “expert” is just a guy with a computer running software that he bought, he has no scientific background in forensics, and he has a “vanity” certification from a board that he runs, that doesn’t exactly make for confidence in his “scientific conclusions.”
Jim H. writes:
This is from Forensic Science Services, Inc., a Canadian company:
The spectrographic voice identification analysis has two steps. The sound of speech is first transformed into a three dimensional (time—frequency—volume) graphic pictures which do reveal numerous acoustical features of an individual’s voice. The second step involves the pattern comparison of the same phrases/sentences from the unknown sample and the suspect’s sample. The results of analysis are expressed as:
The results depend on quality of recordings, the total number of comparison words, speakers’ condition, and individual speakers’ voice variability. There is a requirement for a minimum number of 20 comparison words in a ‘connected speech’. The suspect should provide the comparison sample by reading three times the transcript of the unknown voice sample. [emphasis mine]
- Probably the same speaker (high level of confidence).
- Possibly the same speaker (intermediate level of confidence).
- Inconclusive(due to the insufficient number of comparison words, poor quality of recordings, too high variability of the voice, possible disguise).
- Possibly not same speaker (intermediate level of confidence).
- Probably not the same speaker (high level of confidence).
The EasyVoice software is apparently the “poor man’s” tool for voice analysis. I have been unable to find any evidence that the software has been discussed in any peer-reviewed scientific journals.
Another question: when the software says that there is a “48 percent match” between Zimmerman’s voice and the scream, what does that mean? Forty-eight percent of what? If I have a pile of 100 coins, 48 of which are dimes, and I hand you a dime, then 48 percent of the pile of coins matches what you have in your hand. We know what the 48 percent means.
With Zimmerman’s voice, what does 48 percent mean? 48 percent probability that he was the originator of the scream? That 48 percent of the length of the recording matched his voice? That 48 percent of the tests used confirmed that it was his voice. Does it mean something else? Who knows?
In other words, there is a 48 percent match of SOMETHING between the scream and Zimmerman’s recorded 911 call. And on the basis of THAT we conclude that “science proves” that the scream wasn’t Zimmerman’s?
Slavek B. writes from the Czech Republic:
I have just found an interesting article on forensic analysis by Senior Lecturer at University of Canberra on the issue of forensic voice analysis.
Here is the key part that is relevant to the recently published media results of some experts analysis of the 911 call in Martin/Zimmerman case:
Establishing comparability means being sure that the circumstance of the samples is sufficiently similar for the comparison to make sense. We cannot usefully compare shouting angrily on the phone with speaking calmly to an interviewer.
Those experts mentioned by the MSM were comparing screaming and a calm voice in a phone call: no surprise they were not able to establish positive match. Given the fact how forensic voice analysis work comparing such different samples make no sense at all. It will be extremely easy for the defense lawyer to dismiss such evidence at the court.
Forta Leza writes:
Another problem with this voice analysis is that we do not know how the newspaper found this particular “expert.” Is he the tenth expert that they called? Did the first nine state that there was not enough information to make a determination? Did this Owen person simply volunteer his services to the newspaper? And if he did, was he aware what results were likely to get him the most consideration?
You can bet that if there is a trial, Zimmerman’s side will produce an expert who contradicts Owen. Reasonable people will discount this expert’s testimony since it is known that Zimmerman’s [Martin’s?] attorney has an agenda and that he shopped around for an expert willing to testify to the conclusion he wanted. Similarly, we should discount the testimony of an expert engaged by the news media. And we should also ask why a Florida newspaper is using an expert from New Jersey.
E. writes from Florida:
On a related issue, when was the last time you heard anyone use the term “coon” in reference to a black person? I’m in the South, and I’m guessing I haven’t heard it since the 1960s. I find it hard to believe that a 28 year old Hispanic guy in FL had ever heard the word.
Posted by Lawrence Auster at April 01, 2012 05:04 PM | Send