The integration of human emotions into human-computer interaction has seen significant advancements, yet the subtlety of micro-expressions remains a relatively unexplored area. This paper presents a comprehensive study on the recognition and analysis of micro-expressions using advanced computer vision and deep learning techniques. Utilizing the CASME II dataset, which consists of 247 samples of micro-expressions captured at 200 fps, we preprocess, analyze, and extract features from facial expressions to train a fine-tuned ResNet model. Our methodology includes data augmentation, face cropping, and normalization, followed by the application of TV-L1 Optical Flow Estimation to capture subtle facial changes over time. The results demonstrate the model's robust performance in real-time emotion recognition, with potential applications in gaming, advertising, healthcare, and more. By leveraging a real-time, contactless emotion recognition system, this research aims to enhance the precision and applicability of affective computing technologies.