- Reduce the pain and effort for the customer
- Remove the need to ask security questions
- Wipe out fraud – compare against blacklists and create alerts when suspicious activities are detected
- Reduce call time
How it works?
The technology maps over 50 different physical and behavioral aspects of a voice to produce a unique and independent print.
Your voice is made up of characteristics that are created by the vocal tract, the shape and size of the mouth and nasal passage which generally don’t change over time. Other characteristics are defined by pronunciation, accent, pitch and intonation as well as speed of speech and the emphasis placed on words.
Model Management – the application generates biometric voice prints for each of your customers using a small amount of previously recorded data. Once the audio file is stored in the engine you will be able to automatically recognize your customer.
Streaming – compares audio in real time against one or more voice biometric prints.
For off-line operations, a single CPU-core can process up to 1,000 hours of audio in 24hrs – 42 times real time. Following processing voice biometrics can produce up to 85,000 identifications per second.
The technology has been developed to models channel and speaker variability. Just as your fingerprints are unique to you, your vocal tract contains distinctive characteristics that make it unique and different to that of other people. This technology is intrinsically language independent, you can deploy it in any language without the need for adaptation, changes or configuration.
It can also be used on any communication channel, such as a landline telephone, mobile, VoIP, microphone, satellite, with very high levels of accuracy.
It is also effective over cross channel authentication. It can create a biometric voice print in a specific channel (a landline for instance) and then complete authentications from other, different channels (mobile or VoIP, for example).