Recognising micro-expressions is inherently difficult due to their subtle and fleeting nature. In this work, we present an innovative approach that uses motion magnification combined with transfer learning to analyse microexpression datasets, particularly the SAMM dataset. By amplifying these subtle movements, we transformed micro-expressions into clearer, more distinguishable macro-expressions. Using pre-trained CNNs with frozen layers, we optimized the feature extraction process, achieving a test accuracy of 96. 54%, significantly outperforming previous methods.