Forensic speaker recognition

The aim of forensic speaker recognition is to establish links between individuals and criminal activities, through audio speech recordings. This field is multidisciplinary, combining predominantly phonetics, linguistics, speech signal processing, and forensic statistics. On these bases, expert-based and automatic approaches have been developed to analyze the speaker's utterances on recordings, usually originating from anonymous calls, wiretapping procedures, and covert audio surveillance. Most of the forensic laboratories still opt for either of these two approaches, even though, in many respects, they appear to be complementary. The main requirements for these methods are independence to the text, ability to handle minimal length recordings, and a superior robustness regarding noise, transmission channels, and other variations of the recording conditions. Forensic speaker recognition can be considered a forerunner in the implementation of a logical inference framework to estimate the value of the evidence from the analytical results. The limits of forensic speaker recognition are the absence of a fixed and known number of highly discriminatory features in speech, the limited quality of the audio recordings captured in forensic conditions, and the application of recognition approaches in the absence of any known underlying model that accurately represents the speaker-dependent information.