Explainable and Trustworthy Speech Emotion Recognition Using Confidence Score and Reinforcement Learning Rectified Speech Emotion Descriptors

2026-06-12

AI Summary

This work addresses the limited interpretability and reliability of speech emotion recognition (SER) systems, which stems from the scarcity of datasets annotated with trustworthy speech emotion descriptor (SED) labels. To overcome this challenge, the authors propose an online SED refinement method that integrates confidence scoring with reinforcement learning to post-train and optimize a pre-trained SER system without human intervention. This approach represents the first effort to combine confidence-guided automatic label correction with reinforcement learning, enabling adaptive calibration of interpretable SER systems. Experimental results on the IEMOCAP and MELD datasets demonstrate absolute accuracy improvements of 2.9% and 3.3%, corresponding to relative gains of 3.7% and 5.4%, respectively, thereby validating the method’s effectiveness and practical utility.

Abstract

Explainable and trustworthy speech emotion recognition (SER) remains a challenging task to date, largely due to the scarcity of SER data with reliable speech emotion descriptor (SED) labels, such as prosodic features and speaker traits. This paper presents a confidence score and reinforcement learning (RL) based on-the-fly SED rectification approach for post-training SER systems on automatically annotated SED labels. Experiments on IEMOCAP and MELD suggest that explainable SER systems incorporating the proposed confidence score and RL-based SED rectification approach consistently outperform baselines without data selection or SED rectification. The best performing system, which integrates both components, surpasses the baseline without data selection and SED rectification, achieving SER gains of 2.9% and 3.3% absolute (3.7% and 5.4% relative) on IEMOCAP and MELD benchmarks, respectively.

Problem

• Speech Emotion Recognition
• Explainability
• Trustworthiness
• Speech Emotion Descriptor
• Data Scarcity

Innovation

• confidence score
• reinforcement learning
• speech emotion recognition
• emotion descriptor rectification
• explainable AI