Automated voice identification

Back in the day ... way back when ... Law enforcement needed a way to single-out one person from another. The answer was right at their fingertips. (Ha. Ha.) Apparently we all have unique fingerprints after comparing one to another matching points (curves, intersections, etc.). While my memory isn't all that great I remember my parents taking me to some community building and leaving with ink all over my fingers. Somewhere was a card with a ink blot of the swirls on my skin look like. (Hopefully by now the community has gone digital with this.)

We're talking biometrics. And it has used technology to ... tap ... into areas only [insert Hollywood cliché reference here] or an imposing all-knowing [insert big brother cliché here] could dream of with the various ways to identify someone from another -- fingerprints, DNA, iris scans, my face ... my voice. (There are more.)

Voice biometrics have been around for years; a recording of a spoken phrase is compared to another of the same phrase. Sophisticated computer software breaks down the sound and the two are matched. Having never been in the forensic or an investigative field, this sounds pain-staking. Wasn't technology invented to push the envelope of your efficiency? If you get anything from me other than a few minutes of reading entertainment, please take this tiny nugget of advice: If technology has hindered your work (assuming you were properly trained with/on it), then contact the company. I've spoken to too many manufacturers stating they regularly hear new ideas and suggestions from officers like yourself. Many times these ideas turn into working prototypes. Many times these turn into head-slappingly good ideas.

What Nuance is doing has done

In an interview, Nuance's Brett Beranek, their Solutions Marketing Manager, tells me "it turns out our voice are unique to us" and explained further. There are two sides to your voice: physical and behavioral. In these are multitudes of characteristics; Nuance Forensics' algorithm monitors over 100. Behavioral characteristics include accent, rhythm, speech, etc. Physical traits are affected by a person's larynx, nasal passages, throat, tongue, teeth, etc. 

This intensive algorithm does most of the work - not all, a forensic expert validates it through a handy automatically-created report - or assessment. Later in court the identification of speakers in audio recordings can be confirmed. Before I continue I found it neat that it'll recognize and figure out where I'm from - even if I'm faking my accent. My linguistic nerd, admittedly, became intrigued.

Oh wait. I glossed over something here. Automatically-created? There are even more algorithms at work: gender detection, language recognition, dialect recognition, speaker segmentation (how many people are speaking), etc. Unfortunately no word as to reporting on emotion or deception (yet?). All this once manual data entry has been computerized. If it would only transcribe it too. Even then, forensic experts aren't cut out of the picture - they use this assessment to validate the information for identification. Technology should make us more efficient, not take us out of the picture. Basically this gives that much more data to investigators. 

This report, by the way, is exemplifies my point three paragraphs up; according to Beranek the idea came from working with a forensic expert in law enforcement during the development. 

Have you jumped a few years in the future? If you're thinking that software could soon be able to identify me by my voice alone, know that that isn't the (current) point of biometrics. We're here to confirm Bob's identification ... but ...

National databases, probably more helpful than scary

Fingerprints were once compared paper to paper, smudge to smudge. It was a one to one comparison - in a way it still remains that way. With a national database officers can query it with a fingerprint asking "who is this." The database then compares your data, filters possibilities and provides with matches. As insanely complicated as voice biometrics algorithms sound (a pun!) we're not there yet. The software is comparing one unknown recording to a known other.

However, Beranek says "with continued work with law enforcement community and with continued improvement of the technology itself that it could be used as a very similar fashion as fingerprints are used today. I think we'll eventually get there, maybe five years down the road."

How soon this comes to reality is speculation of course, an exciting - but not fictional - speculation. Back in 2012, posted a brief that states the world had invested such an idea. The report states, "The system, called 'VoiceGrid Nation' can match a recording to a database entry in only five seconds (based on a scan of 10,000 voices), with an accuracy of 90%" This is apparently led by Russia's Speech Technology Center. I wonder how recent tensions have effected this ... has it at all? 

Of course voice technology is just one example of we humans using technology to change our working lives. One only needs a few moments of an over-active imagination to picture how to implement such high-tech concepts into other devices on duty.

What part of your day would you like more -- or less -- technology? How do you see the future ... better yet ... how do you see the future of your duty belt?

Stay safe.


P.S. "Bob" is just a name. All characters appearing in this work are fictitious. Any resemblance to real persons, living or dead, is purely coincidental.