A COMPREHENSIVE METHOD FOR MICRO EXPRESSION RECOGNITION: GRADIENT-FEATURE DIFFERENCE BASED SPOTTING AND AMALGAMATED LBP BASED CLASSIFICATION

Abstract

Abstract: Micro Expression (ME) recognition from video sequences is a challenging task that plays a significant role in various fields, including security, psychology, and human-computer interaction. In this paper, we propose a novel method for ME recognition that encompasses two distinct phases: ME spotting and ME classification. The ME spotting phase identifies key frames by analyzing feature differences based on gradient attributes, capturing significant changes that are indicative of micro-expressions. In the ME classification phase, we employ a combination of Local Binary Patterns (LBP) and Local Average Binary Patterns (LABP) to extract expression-related features from these key frames. These features are then classified using a Support Vector Machine (SVM), which is well-suited for handling complex decision boundaries and achieving high classification accuracy. Our experimental results demonstrate that this approach outperforms existing methods, achieving an accuracy of 72.36% on the CASME II dataset, surpassing previous techniques in both precision and reliability. The proposed method provides a robust and effective framework for ME recognition, offering significant advancements over prior methods and laying the groundwork for future research in this domain.