A Hybrid Gabor Filter-Convolutional Neural Networks Model for Facial Emotion Recognition System

Abstract

Objectives: This research aims to develop a hybrid facial emotion recognition system using integrated machine learning feature extraction techniques with convolutional neural networks model for improving the accuracy of facial emotion recognition (FER). Methods: This study introduces a novel approach that integrates various machine learning feature extraction techniques (HoG, LBP, SIFT and Gabor Filters) with Convolutional Neural Networks (CNNs). The study utilized a set of 10,500 facial emotion images of the CK+48 dataset with six different facial emotion recognition. Findings: The result from the experiments shows that the proposed Gabor Filter-CNN approach for estimating FER had higher accuracy as compared to other CNN, HoG-CNN, LBP-CNN, and SIFT-CNN. The specific approaches is based on the Gabor Filter – CNN paired results with an accuracy rate of 99.85% outstanding other models. Novelty: The proposed approach would ensure the combination of the ML extraction techniques for capturing the details of the textures/features and CNNs for other deep hierarchical features and further better prediction. Keywords: Hybrid Model, Facial expressions, Emotion recognition approach, Machine Learning, Deep Learning, Gabor filter approach, Feature extraction, Convolutional neural network