Exploring the Effects of Device Variability on Forensic Speaker Comparison Using VOCALISE and NFI-FRIDA, A Forensically Realistic Database
|David van der Vloed, Finnian Kelly, Anil Alexander|
In this paper we present NFI-FRIDA (Netherlands Forensic Institute - Forensically Realistic Inter-Device Audio), a database of speech recordings acquired simultaneously by multiple forensically-relevant recording devices, and demonstrate how this database can be used to support forensic speaker comparison (FSC) casework. We use VOCALISE (Voice Comparison and Analysis of the Likelihood of Speech Evidence), an x-vector based automatic speaker recognition system that allows a forensic practitioner to perform speaker comparisons in a flexible way. After establishing how variability of the recording device affects speaker recognition discrimination performance, we explore how variability of the recording device of the relevant population affects the resulting likelihood ratios. These experiments demonstrate a research methodology for how a forensic practitioner can corroborate their subjective judgment of the 'representativeness' of the relevant population in FSC casework.