Facial Expression-Based Emotion Recognition: Analysing Human Affect through Facial Cues

Abstract

This study presents a CNN-RF (Convolutional Neural Network-Random Forest) model for emotion identification using facial expressions. The goal is to create a model that can accurately and efficiently discern emotions from face photographs. For feature extraction and downsampling, the proposed CNN-RF model consists of several convolutional layers followed by max pooling layers. To transform the 2D feature maps to a 1D vector, a flattened layer is employed. The RF layer connects the CNN and RF components and accepts flattened features as input. The completely linked layer connects the RF output to the emotion categorization layer at the end. The experimental findings demonstrate the CNN-RF model's efficacy in emotion recognition. The accuracy values range from 91.69% to 96.44%, indicating a low percentage of false positive predictions, while the recall values range from 92.84% to 96.13%, demonstrating the model's capacity to catch a large number of real positive cases. The F1 scores, which take precision and recall into account, vary from 93.39% to 95.21%, further supporting the model's overall ability to reliably categorize emotions. With an accuracy of 98%, the CNN-RF model demonstrates its capacity to discern emotions based on facial expressions. This study's findings emphasize the CNN-RF model's ability to effectively recognize emotions from facial expressions. The presented model is a viable method for real-world applications where precise emotion identification is critical.