A person's voice is a unique personal characteristic; it is different from all other voices. This is why the biometric method of voice recognition, voice verification, is used for authentication, voice identification, voiceprint and voice signature of persons.
Technically, a biometric voiceprint is captured and stored by the corresponding person as a reference and, during verification, compared with a specific word that the person to be verified must speak. The voice characteristics of individuals differ down to the smallest sound unit, the phoneme, which is expressed in the frequency spectrum of the voice. The voice input is therefore transformed by means of a time-frequency transformation into a frequency spectrum that can consist of several hundred kilobits and thus provides sufficient information for voice recognition.
Voice features are only 1:10,000 in their uniqueness of misclassification. In voice recognition, the number of words stored and interrogated can be reduced. Voice recognition systems are mainly used in telebanking.