By Jared Stern
Fingerprinting has been used for years to determine the individuality of a person. But, newer technology allows investigators to capture a person’s voice, a so-called “voiceprint.” Sometimes, a person’s voice is the only clue that police and forensic teams have to go on.
What Is It?
Voiceprinting is a new kind of forensic evidence that uses a person’s voice to determine certain key facts about a situation or investigation. Voice identification involves the analysis and comparison of a person’s voice, using advanced computer systems and software which can analyze how words flow together, pauses, breath, and the unique patterns generated by each individual’s mouth and larynx.
First, the quality of the recording is established using special instrumentation. In a classic sound spectrograph, the person’s voice is recorded on a magnetic disk and then sent through an amplifier. The sound is then sent through a scanner or frequency analyzer.
The analyzer can separate sounds into different frequencies. From there, a filter selects a group of frequencies and converts them to electrical signals. The signals move a stylus, marking up paper on a recording.
The jagged lines produced by the stylus represent the frequency and intensity (loudness) of the sounds. These two factors – the frequency and intensity – of a person’s voice are what’s important. They are difficult or impossible to fake so long as the recording is verified as authentic.
It’s this voiceprint that interests those in the computer forensic s industry.
For example, in many kidnapping cases, cell phones and computers are used at some point for communication – maybe the kidnapper knows the child or person being abducted and uses the phone or email to contact them.
This leaves digital traces of communication that can later be discovered, analyzed and reconstructed to help investigators locate the suspect and bring him to justice. Even when data appears to have been deleted, investigators can often recover and reassemble it.
Data extraction is another important tool that investigators use when solving crimes. And, with voiceprints, it becomes an even more powerful tool.
According to some new research, private institutions like Chase and Wells Fargo are secretly capturing callers’ voiceprints for use to fight fraud. And customers of financial firm Vanguard access their account by speaking the phrase “At Vanguard, my voice is my password,” into their phone.
These voiceprints are stored on servers and could theoretically be used for other purposes, however, like solving crimes. And it’s something that intelligence agencies and law enforcement are extremely interested in.
How Are Voiceprints Verified?
But like most technologies, there are ways to defeat it. The soundwave, sonogram, and spectrograph must be visually inspected to detect tampering. The recording must be loaded into a computer and tested using separate software that can analyze the characteristics of the recording.
Beyond digital analysis, the forensics expert must manually listen to the voice recording, take notes, and measure and note any differences in pitch, pronunciation, and general speech patterns. There must be a “master recording” with an authentic voice of the person being analyzed for comparison purposes.
Next, the investigator must then take an electronic measurement to measure the spectrum to verify that the individual’s voice is within the same sound range as the “master recording.” Finally, the investigator must view and compare the waveform of the recordings to determine the probability that they are the same.
Finally, the investigator prepares a report for attorneys and the court that can be used as evidence in a trial.
Why Is It Significant?
Voiceprints are similar to fingerprints in that no two are alike. Investigators can use voiceprints to verify the identity of a suspect using complex software capable of analyzing the unique sound patterns generated when a person inhales, creates the vibrations in their “voice box,” and then expels air.
In order for voiceprints to work, the quality of the recording must be acceptable. Sometimes, background noise can interfere with the analysis. The recording must be visually examined, including a visual inspection of the recorder, cassette, or whatever device was used to make the recording.
It’s important that there has been no physical tampering, since a person’s real voice could be inappropriately used to frame them or provide inaccurate or misleading “evidence.”
A visual analysis of the sound wave must be done. Once the recording is visually verified, and no signs of tampering can be found, the recording is copied to a computer and tested.
Finally, a sample from the suspect must be obtained, and the audio characteristics are analyzed. There’s a certain rhythm to how every individual speaks. Even when a person tries to disguise it, the rhythm is still there – enough for a computer program to pick up on. It’s this rhythm that is the “print” of a person’s voice, and the key to uncovering new facts about an investigation.
Jared Stern is the leader of a team of intelligence operatives and special-operations personnel and CEO of Prudential Associates, who focus on crisis management and specialized-force protection solutions. He often appears on radio and television offering expert commentary regarding terrorism, special-warfare tactics and intelligence operations. Look for his posts on various crime and science blogs. Follow Jared on Facebook.
2 thoughts on “Electronic Voiceprints: The Crime Solving Power of Biometric Forensics”
Jared, interesting and informative article. Thanks.
Some earlier basic voiceprint detection systems have been fooled. On the old **Phillips Savvy (1999 model) allowed a user to verbally request a phonebook entry name, having previously recorded a voiceprint. I remember at a conference back in 2000 from the gathering of examiners that were there we were playing around with the Philips Savvy to see if it were possible to make the mobile dial a phonebook entry. Quite a few of our attempts succeeded when each of us tried to replicate the voice of another person; calls were made.
** See page 13 – http://www.virginmobile.com/vm/media/pdf/userguides_qsg/phillips/Savvy/PhilipsSavvyManual.pdf
This issue of mimics (as seen on TV) seems to come to mind when thinking about voiceprint originality; how much published research has there been is to assess the voices of well known personalities (genuine persons) and their mimics?
I digress from what I was actually going to comment on. I do wonder whether voiceprint analysis has a longer historical timeline as opposed to one that maybe said to be contemporary. I see the benefit and link to current voiceprint analysis that has recently been put to good use to assist in the identification of the executed hostages James Foley and Sotloff in 2014. Indeed voiceprint detection was used following 9th September 2011, so I understand.
Nuance product is said to be a conversational technological example of voiceprint used where the target is at a distance (speaking over the telephone lines) is involved. Voiceprint detection is used by banks etc. to analyse previous telephone conversations against a current call, but equally requires to be validated that previous recordings haven’t been changed or tampered with.
However, as a timeline, is it not possible to link voiceprint going back to at least President Richard Nixon and the Watergate tapes of the 1970s? The FBI carried out voice and voiceprint analysis back then, albeit the processes and techniques available to them at that time have improved dramatically today. But the historical link is still there is it not? And shows possibly a 40-years historical timeline for research and development linked to voiceprint and may undeniably add voiceprint as having a qualifying scientific basis that has enabled voiceprint, through refinement of its processes/techniques, to be forensically assessed?
Personally speaking, if this were my field of endeavour I would take comfort from historical links, apart from maybe being able to use an eye-catching headline, “Even a President is not immune to it”.
Apologies that should have read 11th September 2011 and not 9th September 2011. Can this be updated please. Thanks