A study on unveiling hidden factors: Explainable AI for feature boosting in speech emotion recognition

Speech Emotion Recognition (SER) is a pivotal research field that empowers computers to discern human emotions from speech. By identifying emotional cues in vocal signals, machines can engage more effectively in areas such as virtual assistants, healthcare, and customer service. Yet, most SER models remain opaque, functioning as black boxes and eroding user trust, particularly in sensitive settings. Here, we present a SER system rooted in Explainable Artificial Intelligence (XAI). Our system merges real-time audio analysis with transparent machine learning. It extracts key acoustic features from live and recorded speech, then classifies emotions through trained models. Beyond predictions, it illuminates which features most influenced each classification. Experiments show strong accuracy while improving transparency and user confidence. This framework provides a robust and reliable path for building real-time, explainable emotion recognition systems.