Voice identification

Alexander Graham Bell’s father envisioned that a person could be identified by the sound of his or her voice. The concept was to create a visual representation of the spoken word based on subtle differences in pronunciation. In 1941, Bell Telephone in New Jersey produced a sound spectrograph for mapping a voice on a graph by analyzing sound waves based on frequency, intensity, and time. The original purpose was to assist in enabling deaf people to speak. During World War II, the government began using the technology to identify enemy voices on telephones and radios. Currently the National Security Agency uses the software to assist in electronic espionage.
Biometric identification using speech is particularly appealing because it is minimally intrusive. As with fingerprints, a suspect’s voice can be compelled for the purpose of comparison without violating his or her Fifth Amendment rights (see United States v. Dionisio, 410 U.S. 1 (1973)). An individual’s voice contains enough individually unique characteristics to distinguish it from others through analysis. In court situations, the standard for admissibility of an opinion as to the identity of a speaker is merely that the identifier has heard the voice of the alleged speaker at any time. The spectrogram assists the process by graphically displaying sound in three dimensions: pitch, resonating characteristics, and rate of speech.
However, use of the spectrogram presents several challenges. Voiceprints, unlike fingerprints, are not static. Fingerprint ridge details exist from cradle to grave. The human voice, however, changes with factors such as age, stress, and health. Additionally, the process suffers when transmission quality is poor or when ambient noise is present. The accuracy of voice identification depends on the conditions under which the voice samples are made, the characteristics of the equipment used, and the skill of the examiner. Although there is disagreement within the scientific community on the accuracy of voice verification authentication, today’s voiceprint technology has advanced to the degree that several circuit courts now allow comparisons of the spectrogram of an unidentified speaker with that of an identified speaker in order to find similar patterns as admissible scientific evidence.