A Novel Human Face Expression Recognition Based on Image Processing Assisted Convoluted Deep Learning Methodology

Facial expression recognition has become a critical component in applications involving human-computer interaction, security systems, and behavioral analysis. This paper presents a novel approach to human face expression recognition using a convoluted deep learning methodology supported by image processing techniques. The system is designed to accurately classify seven fundamental emotions—Anger, Disgust, Fear, Happy, Sadness, Surprise, and Neutral—using the FER2013 dataset. The methodology includes preprocessing steps such as noise reduction, grayscale conversion, and face registration. Feature extraction is performed using Gabor filters and Local Binary Patterns (LBP), while Convolutional Neural Networks (CNNs) are employed to learn deep hierarchical features. The system was optimized through hyperparameter tuning, achieving an overall accuracy of 96.29% on the test set. Precision, recall, and F1-scores for each emotion also exceeded 95%, with “Happy” and “Surprise” emotions showing the best performance. The results indicate that the model is effective for real-time applications requiring reliable emotion detection. The paper concludes with discussions on model performance, challenges in classifying similar emotions, and recommendations for future improvements.