Forensic voice comparison (FVC) involves a comparative analysis of suspect and offender speech data for use in a court of law, with the Bayesian Likelihood-ratio (LR) framework being one common approach for undertaking this task. Central to this approach is making comparisons against a large background of speakers with similar speaking styles as the offender in question. But often not all of the speech data to be compared has been acquired at the same sampling rate, thus necessitating sampling rate adjustments of some kind. This project will investigate which sampling rate approach, namely up-sampling or down-sampling, gives better FVC performance.
Lab allocations have not been finalised