Friday, 05 Dec, 2008 Science

Voice Analysis Cannot Identify Someone With Certainty on Its Own


According to an expert from the University of New York, determining someone from the sounds they produce when speaking (which often can be seen in popular crime shows) is not as easy as one may think. Thus people get an unrealistic idea of what science can really do.

"Most dramas include this kind of thing [and] there's a great deal of poetic license in what they do," says linguistics and phonetics expert Dr Paul Foulkes about speech analysis. It is worth mentioning that the American scientist previously provided training in forensic speech analysis exclusively for the US Secret Service and the FBI. Currently he works as a freelance advisor for a private forensic speech science lab.

At the conference entitled ARC Network in Human Communication Science, which took place in Sydney, Foulkes stated that scientists using forensic speech science, which includes acoustic, phonetic and linguistic analysis of recordings, may reduce the number of possible suspects. In addition, it may be used to identify how close the voice of a suspect to that of a criminal is. But at the same time Foulkes outlined that a recording of a speech cannot determine someone with certainty on its own.

"There is no such thing as a voice print. It's a very very dangerous term. There is no single feature of a voice that is indelible that works like a fingerprint does," he says.

Foulkes mentioned that there are a lot of various factors that influence the way a person speaks in a specific time and place. The voice of a person may sound differently if he or she is, for example, speaking on a cell phone in traffic or if a person has a cold. Thus one cannot use the features of speech and language in order to determine the identity of one individual, excluding other people under regular circumstances. "People's voices overlap," he says.

Although speech analysis in reality cannot determine someone on its own, it may be useful for eliminating certain suspects. Such voice features as pitch, pronunciation of vowels and consonant and the quality of voice can be analyzed with the help of phonetic and linguistic analysis. Foulkes mentioned that the forensic investigator may also determine some pathological features of voice, such as the pronunciation of "r"s as "w"s, reports ABC News

Foulkes expressed his concerns regarding the use of software to analyze the acoustic features in a person's voice to identify their state of mind. The expert outlined that "lie detectors" or so-called "voice stress analyzers" are mainly based on "questionable principles." Nevertheless, some organizations use them for security reasons.

"Moscow airport is using this as a way of detecting whether someone is a terrorist threat or not. I think that is a very worrying development," says Foulkes.

Powered by


46 votes

//2 Apr 15, 2010 02:03 AM | posted by: Paul Foulkes
Just to clarify that I'm British, not American, and at the University of (Old) York, UK. In response to Cynthia's question - in general the answer is asymmetrical, and depends on the context. It is asymmetrical in that it is easier to eliminate someone from consideration (ie give a negative answer) than say for certain it IS a particular speaker. The latter is only possible if there is a finite (and small) set of candidates. For example, if it is known from video surveillance that 3 people are in a car in which a bug has been placed, then it might be possible to say pretty certainly which of the 3 spoke which words. Please drop me a line if you want to talk more about such topics.
43 votes

//1 Mar 19, 2010 01:49 AM | posted by: Cynthia Medley
is there anyone that can do a voice analysis from a recorder, and prove that it is or is not a certain person.

Add your comment:

antispam code





Discover, share, comment and discuss with us on a variety of interesting stories. A lot of fascinating things are taking place every day around the globe and we welcome you to this world.