All of us have little difficulty identifying people we know by their voices. But to prove that a voice belongs to a person and thereby use it as admissible evidence in court calls for precision of technology and a high degree of skill. This is where forensic voice analysis comes into picture.
It was Melville Bell, father of Alexander Graham Bell, who first gave a visual representation of spoken words over a century ago. The method using handwritten symbols conveyed much information about pronunciation and was used to teach the deaf to speak. Bell laboratories in the 40's did pioneering work in the development of sound spectrograph. A spectrograph acts as an automatic wave analyser and records speech sounds across the three domains of frequency, intensity and time. During WW-II acoustic scientists used it to identify enemy voices in intercepted radio signals. But it was only during the 60's that voice prints began to have forensic application.
The foundation for forensic voice identification is the premise that like finger prints, each individual voice print is also unique. Two factors contribute to this uniqueness. One, each individual's vocal tract is different in shape and size. The vocal tract with its oral and nasal cavities acts as a resonator of the fundamental tone produced by the paired vocal cords. Selective processing of the sound, that is, voice, takes place depending on the size and manner in which the cavities are coupled. Two, the way a person moves the articulators involved in speech, such as the tongue, jaw muscles, lips, etc. in dynamic coordination is also highly individualistic. The chances that two individuals share identical vocal tract dimensions and configurations as also the manner of producing speech are so remote that personal identification by voice is possible.
A voice print identification can be made when there are common speech sounds in both the known and unknown speech samples.
Forensic Science laboratories such as the one in Chandigarh, use voice identification techniques regularly and the Supreme Court has held that voice identification data is admissible in court. In Bangalore, SRC Institute of Speech and Hearing has the facilities for voice analysis. The All India Institute of Speech and Hearing, Mysore, which has been working in the field for many years now, even wants to start a one-year PG Diploma course in forensic voice analysis.
There are softwares which can compare particular parameters and show the extent of match. With spectrograms, an expert looks at the fundamental frequency of the voice, formats, etc. and side by side visual comparison of speech sounds is made. The perceptual analysis task requires an examiner to hear two speech samples, one after another, and rely on memory to compare the two acoustic samples. However, in aural analysis an expert also listens to many other parameters such as breathing patterns and accent while making the identification. Examiners usually listen to and compare several recordings of the speaker's voice. With combined aural and visual analyses of samples, examiners reach conclusions such as positive identification, probable identification, positive elimination, probable elimination or no decision.
Background noise, intra-subject variability where a speaker's voice changes due to reasons like health/emotion/age, situations when a person deliberately disguises his voice are confounding factors in voice identification. However, a thorough knowledge of voice science and linguistic features, lengthier speech samples and experience help. Experts point out that age related changes are gradual and because several parameters are compared deception is not easy.
Forensic voice identification has now become a sophisticated and standardised method, making use of the latest in audio recording and computer technology. But opinions are still divided over the fool proof nature of the procedure. With greater sophistication in technology and more stringent standardisation, the sceptics are likely to be won over.
