ARVMex: A web application for obtaining acoustic reference values for forensic studies in Spanish

Abstract

Forensic speech experts using the auditory-acoustic-phonetic approach face different challenges when dealing with recordings collected under different technical conditions. We developed ARVMex tool to provide acoustic reference values to describe variation due to technical conditions in recordings. The tool employs a corpus of Spanish speakers, generating descriptive statistics based on different attributes, including speaker characteristics, different sounds, and on two technical conditions: face-to-face interviews and telephone conditions. The results indicate that, for the fundamental frequency, the mean values of the interview attribute accurately depict the behavior of each vowel by gender, whereas the mean values of the telephone attribute exhibit deviation from the general average. In relation to vowel formants, reference values per vowel, gender, and technical conditions must be established considering variations among vowels and the overall mean. The tool presented here aims to assist experts by providing insights into the variations in the measurements of acoustic properties due to technical conditions.