Identificação de falante: um estudo perceptivo da qualidade de voz

Abstract

Introduction: the identification of speakers for forensic means started at the beginning of the eighteenth century. Nevertheless, the standardization of procedures and the use of a single voice recognition protocol still seem to be far from achieving. Thus, it is necessary to ensure the real contribution from perceptual analysis of voice quality to the identification of speakers. Objective: to check the applicability of the script Vocal Profile Analysis Scheme (VPAS-PB) in the identification of speakers, from the results of a perceptive experiment based on the phonetic description of voice quality and dynamic model. Methodology: the corpus is composed by semi spontaneous speech samples taken from the reading of a sentence and the telling of two personal narratives with important facts from the lives of ten male subjects. The samples were recorded simultaneously by a voice recorder and a cellphone. Two groups of judges of the VPAS-PB script have participated in the research. The first group, composed by three experienced judges in the use of the instrument, heard the speech samples and applied the script. The second group of judges, composed by eighteen academics in the Phonoaudiology area, participated in a perceptive VPAS-PB training course, and the individuals in the group were assigned the task of listening to five speech samples and matching each of them to the description of it. They were given fifteen audio files in the extension .way, with 5 speech samples each (lineups) and a sheet of paper containing fifteen voice quality descriptions, which they had to hear, analyze and match. The descriptions were assembled from the average patterns stated by the first group of judges. Results: the statistical analysis has shown a high reliability of experienced judges in the application of the VPAS-PB script. The results of the perceptive task performed after training have indicated that, judges in training rely primarily on the strategy of using aspects of the vocal dynamic for vocal quality identification. Conclusions: the VPAS-PB script can be a useful complementary tool for speakers identification, once it presents a high degree of reliability by experienced judges. The perceptive training carried out has shown the applicability of teaching how to manage the instrument, proving it to be vitally important in the training of experts in speakers identification. The experience with the training and the results found provide us with subsides to improve the strategies in perceptive training in the future