Known as a spectrogram, a visual representation of the voice is created on a computer. This voiceprint captures the vibration frequency (pitch), amplitude (volume) and time of a person's vocal cords. It recognizes three types of sounds: voiced sounds (vowels); unvoiced sounds ("sh"); and plosive sounds, which occur when pressure builds in the mouth and is released.
The voiceprint is not a recording or the voice's wave pattern and cannot be played back. It is a digital interpretation of a person's voice and must be processed for comparison by a complex algorithm. Some algorithms employ more than a hundred calculations.
Two ways of comparing voiceprints exist. The text-dependent approach examines a pair of voiceprints of identical phrases or words. The text-independent approach does not use the same words. Instead, similar patterns within the words are identified. The latter comparison takes more time.