THE TECHNOLOGY OF VOICE BIOMETRICS
By Iain Robinson, Director, Sales & Business Development Northern Europe, Agnitio
“If somebody calls you and you don’t recognise the voice – all you know is that somebody called.”
Today terrorist and criminals hide behind the anonymity provided by technology or the simple pay as you go phone, where no contract is required and phones have almost become a use once and throw away item.
In today’s world of global terrorism and serious organised crime a new way of identifying targets is required. The technology of Voice Biometrics provides a means to add value to forensic evidence, intelligence gathering and situational awareness. It provides a means to identify, overtly or covertly, a speaker and to associate the telephone and other records to the voice to build the intelligence picture.
For those who don’t know Voice Biometrics is the representation of the Physiological information embedded into voice waves coming from your vocal tract. This means that the air forced through the throat from the vocal chords and modified by the nasal cavities, tongue and teeth provide a signature that is unique. They are not modified by your emotional state, the language you use or by colds. If you try to mask your voice by covering your mouth with a cloth the Agnitio technology has the means to detect this and the call is tagged for further review ; additionally if you try to play a recorded voice this can also be detected and the call tagged for further review.
For these reasons Voice Biometrics can identify you no matter what language you speak, or channel (GSM, PSTN, Satellite, and VoIP) you use. In real life conditions, voice has proven to be as or more reliable than fingerprints or iris scans and a viable technology for operational deployments.
Voice Biometrics can be used to answer the simple questions of, “Who said that?” and “Is that X speaking?”
You could analyze millions of incoming calls (continuous flow) spotting a list of valuable targets for intelligence and surveillance purpose. You could store them in a centrally managed database of all the voice samples and models of your suspects (and search it). You could distribute them to all your speaker surveillance solutions. You could support special operations providing on the spot identification and allowing immediate action when in the field i.e. Asset tracking using a UAV.
There is more that can be achieved when Voice biometrics are included in Command and Control or integrated intelligence systems. Instead of knowing only which telephone has “called” and which phone has been “called” you can start to identify:
* how many phones a suspect has
* track people using their voices; if you know the geographical location of the phone
* Find clusters of calls with the same speaker.
* Add voice data to behavioral analysis and risk assessment applications
Of course all that has been mentioned so far is already being done by somebody somewhere, but still more is available. Can you imagine all this in an area of operations, supporting C-IED operations, providing voice checks at control posts using handheld devices to help identify the innocent population and spot the terrorist, this is also available in some countries where other biometrics are also stored.
So what do you need in a Voice Biometric, speaker identification solution?
Firstly some means to intercept and record the audio from a call; this may be a mass intercept system, an exchange “butinski” type device, a simple application on a laptop or a covert microphone. You will need to record the audio in a suitable file format and store it in a simple database with some Meta data. Alternatively systems are available that will allow audio streaming and identification. Any information on the target, location, time and picture would help future analysis. A communications channel to pass data from an operational area may also be required but not essential as this depend