Acoustic signal analysis for forensic applications

Acoustical analysis of audio signals is important in many legal contexts for determining the authenticity, originality, and continuity of recorded media; determining the circumstances of events in question that may have been recorded; for determining the audibility of signals; and for identification or elimination of talkers as a match to an unknown exemplar. Recorded media are analyzed in forensic applications using both familiar techniques (waveform and spectral analyses) and more novel methods (e.g., ferro fluid development of media; specialized tape heads with nonstandard reproduction characteristics; crystal microscopy; detection and matching to power grid frequencies). Audibility analyses frequently require careful reconstructive field measurements and criteria in excess of normally accepted standards. Voice identification‐elimination protocols must account for examiner bias and exemplar quality and can be described using a receiver operator curve (ROC) model. This presentation gives an overview of these techniques and their comparative advantages for specific forensic applications.