Speaker and prosodic peculiarity classification in emotional speech

Résumé fourni par la source

In this study, the relationship between rhythmic metrics, emotion recognition, and speaker variability is investigated using the German emotional speech corpus (VMEmo). Using principal component analysis and linear discriminant, the results show accuracies close to 0.40 when rhythmic features from different acoustic domains of time, intensity, and frequency are merged to identify linguistic behavior. However, the fluctuating accuracies of 0.44 to 0.17 in classifying speakers based on specific rhythmic feature categories emphasize the significant differences within these feature subgroups. These variations suggest possible nuances or complexities that require deeper exploration and thorough investigation to better understand the differences between these features and their impact on speaker classification accuracy.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.