Facial Emotion Recognition (FER) plays a vital role in various domains, including security, healthcare, and human-computer interaction. While traditional FER focuses on macro expressions, incorporating micro expressions—brief, involuntary facial movements—can significantly improve accuracy and reliability. This paper analyzes macro and micro expressions, along with explainability, to advance FER. We explore the challenges of micro expression detection, including limited datasets, and discuss the advantages of advanced models like Capsule Networks for capturing nuanced facial features. The importance of explainable AI for transparency and trust in FER is also addressed. Specifically, we introduce a Capsule Network-based model for emotion recognition, analyzing both macro and micro expressions. Explainability is integrated using LIME and Grad-CAM to offer an understanding of the model’s decision-making process. By combining advanced modeling and explainability, we propose a framework for robust and trustworthy FER systems, aiming to improve both performance and user acceptance for broader applications. Our model demonstrates strong performance on datasets with clear expressions but encounters challenges with complex and imbalanced data, highlighting areas for further improvement.