Accessibility statement

What is forensic speech science?

Forensic speech science is the application of linguistics, phonetics and acoustics to legal investigations and proceedings.

Speaker comparison

Speaker comparison is the most common task in forensic speech analysis. It involves the comparison of speech patterns to assess the possibility of identity or non-identity of voices. Comparison is made by examining the phonetic, linguistic and acoustic components of the speech, and may be supplemented by automated procedures using speaker recognition technology.

Voices are highly variable, and comparison can be a very complex matter.

Applications

Disputed utterance analysis and enhancement

Disputed utterance examination involves analysis of problematic sections of recordings to determine what was said. There may be dispute, for example, because of noisy conditions, overlapping speech, accents, dialects, or even language.

Enhancement may be carried out to assist in this kind of work. Enhancement involves applying digital filters and dynamic processors to recordings to reduce background noise and improve intelligibility of speech.

Applications

Speaker profiling

Profiling involves analysis of recordings of unknown speakers to gain information about the regional and social background of the speaker.

Background noise may also provide information about where and when the recording was made.

Applications

Voice line-ups

Sometimes voice evidence can be crucial in a case even when it has not been recorded. A witness may have heard a voice but not seen the face of the speaker, for example if receiving abusive phone calls, witnessing a crime in the dark, or encountering masked robbers.

It is now generally accepted that witnesses must undergo formal testing in order to demonstrate their ability to recognise a particular voice. This is usually done via the construction of a voice "identification parade" in which the witness is asked to identify the criminal's voice within a group of foils.

Forensic speech analysts are also involved in general experimental work to test recognition and recall of voices.

Applications

Authentication

Authentication involves examining recordings for evidence of editing or tampering, or assessing whether a recording has been made on a particular machine.

Applications

Forensic audio analysis

Some forensic cases involve analysis of recordings of non-speech sounds such as gun shots, machine noise or coughs.

In other cases sound propagation tests may be carried out, for example to assess whether a particular sound would have been audible to a witness at a particular location.

Applications