In classic whodunnit mysteries, detectives and FBI agents dust for fingerprints to solve mysteries and collect court-admissible evidence.
In real life, it's more often the voice that offers the tell-tale evidence, since technology to recognize voices in recordings has become so much more sophisticated.
The Feb. 26 recording of a 911 call by a woman who reported someone crying out for help in her gated community in Sanford, Fla., could be a key piece of evidence in the Trayvon Martin murder case, especially since she called early enough so that screams for help and the gunshot were recorded.
George Zimmerman, a Neighborhood Watch volunteer, shot Martin, an unarmed 17-year-old. Why he did so remains a hotly argued topic, with Zimmerman claiming Martin attacked and beat him.
Tom Owen, a forensic consultant for Owen Forensic Services LLC and chair emeritus for the American Board of Recorded Evidence, recently analyzed the tape. After running the woman's 911 call through a software program called Easy Voice Biometrics and comparing it to another 911 call with Zimmerman's voice, Owen's team concluded that the screams for help were not Zimmerman's.
And, Owen said, if he had samples of Martin's voice, he may be able to definitively identify the screams as his.
"We've talked to the family; the attorney has been notified," he told Discovery News.
How can he be so sure?
Acoustic scientists have been using audio forensics since WWII when, with the aid of the newly invented sound spectrograph, they realized they might be able to identify enemy voices on radio broadcasts. The spectrograph graphed the frequency and amplitude of voice patterns.
The scientists realized the value of that information because, with the possible exception of twins, voiceprints are unique, forensic-audio examiner Stuart Allen told Discovery News.
"Unless you've had surgery, you can't disguise the characteristics of your vocal cords or mouth structure," he said.
The technology has improved vastly since then: today's software can show the actual speech structure of the vocal cords, Allen said. Although it's used by forensic experts, one of the software programs was created for speech experts to invent therapeutic remedies for people with congenital deformities such as a cleft palate.
In a typical case, Allen takes several phrases of at least 10 words from a wiretap or telephone call and generates a voiceprint with a software program by a Speech Technology Center called Trawl. It's a program so sophisticated that government agencies worldwide use it to analyze phone calls in real time by comparing voices to a database of 10,000 known voices.
"The call comes in and scans the voice and says who it is," Allen said.
Allen makes a transcript of the phrases he's analyzed, then gets the defendant to read it.
"In most cases, defendants try to play with it and disguise their voice," he said. "But no matter how they try, I gotcha ... I compare them and line up the similar characteristics in the exact words. At the end of the day, there are seven outcomes, from inconclusive to a high degree of scientific certainty."
The software spits out values on each element of speech, including pitch, energy distributions, word length (the time it takes to say a word), coupling (the nasal characteristics of the speaker), intonation and emphasis on certain sounds or words.
"The probability of a match must exceed 90 percent for me to call it within any degree of scientific certainty," Allen said. "A perfect match to an identical set of knowns is 99.9 percent."
The program the Sentinel's experts used, Easy Voice Biometrics, is easier to use than Trawl, Allen said. When he testifies in court, Allen said he uses two programs.
Of course, this case is different: a media organization requested the analysis. Allen said he would have recused himself from that situation, worrying that it could taint court proceedings.
Not everyone claiming to be an expert uses software. Ed Primeau told Discovery News he uses critical listening, like a piano tuner.
"The human voice is like a symphony, each voice is unique based on the type of instruments played and skill of the musicians playing the instruments. Similarly, the human voice is created using many physical components, the lungs, larynx and wind pipe. These exit the body through the mouth involving our tongue, teeth and lips," Primeau said.
"As forensic experts, we do not have the proper tools to arrive at a positive ID to conclude the voice is Trayvon yelling for help," Primeau said. "By process of elimination, both Tom and I agree the voice yelling is not Zimmerman. That is our opinion."
Trayvon Martin's brother has said that the screams on the tape sound like his brother.