Facial Expression Recognition Enhancement Using Convolutional Neural Network

Facial expression recognition (FER) technology is quite popular in the fields of computer vision, security monitoring, image classification, and many other related applications. Enhancing the computer's ability to read facial expressions is important for human-computer interaction, as it enables machines to understand and interact with human emotions. In this paper, a modified approach using neural networks (CNNs) is presented to accurately identify unique facial expressions. using the changes to a commonly used 12-layer CNN model to improve its performance in facial expression recognition (FER). The model is trained using dataset images, which allows it to better infer people's emotions from their facial features. To enhance the accuracy of the system, a preprocessing stage is incorporated. This stage involves several operations for data augmentation, including changing the color in images, such as HSV (Hue, Saturation, Value), YCbCr (Luma, Blue-difference Chroma, Red-difference Chroma), etc., to facilitate better interpretation of color information. Additionally, the preprocessing stage refines facial expression recognition by manipulating facial features and adding extra RGB details, improving the visual information provided by the images. The study specifically focuses on evaluating the effectiveness of this approach using the KDEF database. The KDEF database contains standardized images of facial expressions, making it suitable for assessing the performance of the proposed system. By combining the modified CNN model, training with images, and the preprocessing operations, the system's performance was significantly enhanced. As a result, it achieved recognition rates of up to 95%, indicating a notable improvement in accurately identifying unique facial expressions compared to previous approaches.