The ability to automatically recognize posed and spontaneous facial expressions is crucial for affective computing, human-computer interaction, and psychological analysis. This complex task requires identifying both the type and authenticity of emotions, providing valuable insights into human emotional states. In this study, we frame the problem as a 12-class classification task, distinguishing six universal emotions between posed and spontaneous expressions. Instead of processing entire video sequences, we introduce a computationally efficient approach that selects a single “key emotion frame” to represent an expression. Using the PEDFE dataset, designed by psychology experts, our method achieved 68.45% accuracy, significantly surpassing expert evaluations (49.42%) and previous deep learning studies. These results demonstrate the effectiveness of our approach in improving accuracy while reducing computational complexity, advancing emotion recognition research.