Automatic Peak Frame Selection from Dynamic Facial Expressions

Studies on facial expression recognition have been active in recognizing human emotions by analyzing faces since humans express their emotions through facial expressions. In the real world, the time for expressing facial expressions is short and facial expressions are drastically changed over time. Capturing that brief moment is vital for recognizing emotions since it provides valuable and relevant information. Similarly, capturing the peak moment that provides a fully expressive facial expression for the respective emotion is essential for training the facial expressions recognition model. This paper proposes the methodology to automatically spot peak frames, which provide the most expressive facial expressions from image sequences. Our experiment will utilize the Multimedia Understanding Group (MUG) Facial Expression database to conduct our proposed methodology. Our method achieves 95% accuracy performance with more simplicity and time-efficient.