Paraconsistent Feature Analysis for the Competency Evaluation of Voice Impersonation

Mimicry is the art of imitating public figures, animals, and instruments. It’s a popular profession in entertainment, where skilled individuals synchronize voice and gestures, but it is also used deceptively in voice-based crimes. We conducted a paraconsistent feature analysis (PFE) on acoustic features to gain a better understanding of their role in detecting highly competent voice impersonators. To validate our observations from the feature analysis, we performed an experiment on a mimicry dataset using four classifiers. Our proposed system evaluates the competence of artists in voice mimicking and ranks them based on scores obtained from a classifier. A ’hit’ is recorded when the system identifies the artist with the highest mean opinion score as rank-1. Performance is evaluated using the top X-hit criteria on the mimicry dataset. Through our experiments, we demonstrate the potential of PFE in selecting optimal features for assessing the best voice-mimicking candidate.