A human voice is as distinct as a human's face image or fingerprints. The popularity of diverse communication means (landline and cell phones, Internet phones, etc.) open a wide number of opportunities for the application of voice recognition technology. In addition, identifying humans by their voices is both user-friendly and effortless.

Voice recognition technology uses distinctive voice characteristics to verify a person's identity. These characteristics are based on physiological and behavioral components.

It should be noted that the voice (alongside with handwriting and gait) is a behavioral feature, resulting in its low reliability and accuracy.

Voice recognition techniques are used in a number of application areas related to processing user requests by phone (such as call-centers), which allows faster servicing of clients and reduced operator workload. In high-value projects (such as those designed to secure confidential data) voice recognition is good at supplementing other biometric technologies (mostly, fingerprint biometrics).

Voice Biometric Features

Handwriting recognition is based on analyzing such distinct physiological components as the physical shape of the mouth and throat, structure of the vocal cords as well as such acquired habits as the volume, manner and rate of speech.

The voice is highly dependent on emotional factors (a person's mood) and state of health (sore throat, running nose or bronchitis). Identification quality is also governed by a number of external factors (traffic noise, other people talking, etc.). If communication lines are used to identify a human by voice, background noise can also make it difficult to perform voice identification.

Processing Voice Samples

Human speech is divided into several segments, composed of several dominant frequencies called formants. These segments are then digitized, and the biometric templates, the so-called “voice prints”, are generated. For identification, the previously enrolled voice prints and newly produced ones are compared for matching.

To increase reliability and accelerate identification, humans are often asked to answer assigned questions or speak a password or a fixed pass phrase. In this case, humans are recognized in the verification mode.

BioLink solutions based on voice recognition are:

• BioLink AMIS, an automated multi-biometric information system.