MGCP-FER: Meta Control Genetic Programming (MCGP-FER) based feature extraction and fusion for facial expression recognition.

The high significance of the human-machine interaction system, facial expression recognition (FER) has become one of the most popular research fields in pattern recognition. The recognition of facial expressions is a challenging problem, because it is dependent on the quality of the facial depiction. Many different aspects have been added to detect facial expressions. But there is no universal set of features that can be useful to capture all the emotions. Therefore, there have been some researchers who are interested in combining several features to get more recognition. But this method is prone to high-dimensional problems of combined characteristics and redundant information. To address these problems, a new facial expression recognition system is proposed, which consists of Meta Control Genetic Programming (MCGP) for feature selection and fusion in the FER framework. The proposed system consists of a series of processes, including preprocessing, feature extraction, feature selection/fusion, and recognition. The initial phase of the analysis is the preprocessing of face images to suit the further analysis. After preprocessing, the system extracts three different features, namely Enhanced Local Binary Pattern (ELBP), Histogram of Oriented Gradients (HOG), and Geometric Feature (GF). These features are critical for capturing texture and shape features. The MCGP is employed to select and combine the most relevant characteristics to reduce redundancy and dimensionality. Finally, Support Vector Machine (SVM) is employed to recognize expression. The results obtained from two facial databases, such as CK + and JAFFE, demonstrate that the proposed FER system surpasses benchmark models across both databases. This shows that in addition to increasing the identification rate, the method also contributes to a more efficient and reliable approach to FER.