Explainable AI for Speech Emotion Recognition

Abstract

Artificial Intelligence (AI) has become essential across domains, excelling in classification, regression, clustering, and optimization [1]. However, the opacity of traditional AI models, particularly in Speech Emotion Recognition (SER), highlights the need for greater explainability [1]. This research advances Explainable AI (XAI) by developing SER models [2], [3]. It integrates insights from a Literature Review, enhances human-centered XAI methods, and utilizes 18 features for analysis. A feature range metric assesses model performance and explanation quality [4], contributing to a more transparent and interpretable AI framework for SER.