This article is from the source 'bbc' and was first published or seen on . It will not be checked again for changes.

You can find the current article at its original source at http://news.bbc.co.uk/go/rss/-/1/hi/magazine/5402832.stm

The article has changed 2 times. There is an RSS feed of changes available.

Version 0 Version 1
How accurate is voice identification? How accurate is voice identification?
(10 minutes later)
WHO, WHAT, WHY? The Magazine answers...WHO, WHAT, WHY? The Magazine answers...
Sheridan says the tape is fakedThe News of the World claims it has a videotape which proves Scottish politician Tommy Sheridan lied under oath, because experts have identified his voice admitting using a sex club. But how certain can they be it's him? Sheridan says the tape is fakedThe News of the World claims it has a videotape which proves Scottish politician Tommy Sheridan lied under oath, because experts have identified his voice saying he used a sex club. But how certain can they be it's him?
"A stitch-up" is how Mr Sheridan, the former Scottish Socialist Party leader, has branded a tape which allegedly shows him confessing to attending a swingers club."A stitch-up" is how Mr Sheridan, the former Scottish Socialist Party leader, has branded a tape which allegedly shows him confessing to attending a swingers club.
In August, the Glasgow MSP won his defamation case against the paper after it accused him of cheating on his wife and attending orgies.In August, the Glasgow MSP won his defamation case against the paper after it accused him of cheating on his wife and attending orgies.
On the new tape, the face of the speaker is not visible but the News of the World says four voice analysts, including one who worked on the "Osama bin Laden" tapes, have identified him as Mr Sheridan. The politician has accused the paper of manufacturing the voice.On the new tape, the face of the speaker is not visible but the News of the World says four voice analysts, including one who worked on the "Osama bin Laden" tapes, have identified him as Mr Sheridan. The politician has accused the paper of manufacturing the voice.
So how do experts identify speech and how accurate can they be?So how do experts identify speech and how accurate can they be?
WHO, WHAT, WHY? A regular feature in the BBC News Magazine - aiming to answer some of the questions behind the headlinesWHO, WHAT, WHY? A regular feature in the BBC News Magazine - aiming to answer some of the questions behind the headlines
Forensic speech scientist Peter French says voice identification by an expert takes two forms - phonetic and acoustic testing.Forensic speech scientist Peter French says voice identification by an expert takes two forms - phonetic and acoustic testing.
A highly trained ear can listen to a voice and perform a phonetic test, which involves transcribing the speech using a system of symbols to denote how, for example, the person pronounces consonants and vowels.A highly trained ear can listen to a voice and perform a phonetic test, which involves transcribing the speech using a system of symbols to denote how, for example, the person pronounces consonants and vowels.
The tests also assess other aspects of speech, such as rhythm, intonation, voice quality and resonance.The tests also assess other aspects of speech, such as rhythm, intonation, voice quality and resonance.
The second form of identification, acoustic analysis, is carried out by computer programmes which measure a range of voice parameters such as the rate of vocal chord vibrations, which produce the pitch.The second form of identification, acoustic analysis, is carried out by computer programmes which measure a range of voice parameters such as the rate of vocal chord vibrations, which produce the pitch.
SmokingSmoking
"There's a large measure of science in it and a bit of subjective judgement too," says Mr French, a leading figure in his field."There's a large measure of science in it and a bit of subjective judgement too," says Mr French, a leading figure in his field.
"When you look at the results you find they never exactly match 100%. It's not a fixed biometric like a fingerprint."When you look at the results you find they never exactly match 100%. It's not a fixed biometric like a fingerprint.
"Speech is a reflexive of human activity and moving your lips and vocal chords can vary, just as people may move their legs differently at different times."Speech is a reflexive of human activity and moving your lips and vocal chords can vary, just as people may move their legs differently at different times.
Experts use the International Phonetic AlphabetExperts use the International Phonetic Alphabet
"It's affected by things such as tiredness, emotions, smoking and drinking. People also move up and downmarket according to who they are speaking to.""It's affected by things such as tiredness, emotions, smoking and drinking. People also move up and downmarket according to who they are speaking to."
This means the results need to be looked at very carefully, he says, to consider whether the differences in results are inconsequential or whether they indicate this could be two different people.This means the results need to be looked at very carefully, he says, to consider whether the differences in results are inconsequential or whether they indicate this could be two different people.
Experts should never say conclusively they have identified a person and this kind of evidence should never solely be used to bring a criminal trial, he says. But he added he had never heard of a voice being artificially created to match a real one.Experts should never say conclusively they have identified a person and this kind of evidence should never solely be used to bring a criminal trial, he says. But he added he had never heard of a voice being artificially created to match a real one.
"Experts can never be 100%, it's a balance of probabilities," says voice expert David Howard."Experts can never be 100%, it's a balance of probabilities," says voice expert David Howard.
An archetypal Yorkshire or Lancashire accent may sound fine in the theatre in the evening but it won't be enough David HowardVoice expertAn archetypal Yorkshire or Lancashire accent may sound fine in the theatre in the evening but it won't be enough David HowardVoice expert
"They can give an opinion which leans one way or the other but never 100%."They can give an opinion which leans one way or the other but never 100%.
"There's no such thing as a voiceprint like a fingerprint. So they can't say unambiguously 'That's Mr Smith'.""There's no such thing as a voiceprint like a fingerprint. So they can't say unambiguously 'That's Mr Smith'."
But there could be a way of pronouncing vowels or other acoustic pointers which make the experts believe it more likely the person on the tape is Mr Smith.But there could be a way of pronouncing vowels or other acoustic pointers which make the experts believe it more likely the person on the tape is Mr Smith.
And mimicry is unlikely to trick the experts, he added.And mimicry is unlikely to trick the experts, he added.
"If you're going to change your voice, you have to be really rigorous in changing all aspects of it. An archetypal Yorkshire or Lancashire accent may sound fine in the theatre in the evening but it won't be enough.""If you're going to change your voice, you have to be really rigorous in changing all aspects of it. An archetypal Yorkshire or Lancashire accent may sound fine in the theatre in the evening but it won't be enough."