Investigating Dense Cnn Architectures: A Case Study of Facial Emotional Recognition Systems

Investigating Dense Cnn Architectures: A Case Study of Facial Emotional Recognition Systems

Emmanuel Kyei, Peter Appiahene, Mighty Abra Ayidzoe, Obed Appiah, Justice Asare, Emmanuel Freeman, William Brown-Acquaye

In recent years, Facial Emotional Recognition (FER) has garnered significant attention for its pivotal role across various applications, including human-computer interaction, healthcare, and sentiment analysis. This study is motivated by the need to enhance the accuracy and efficiency of FER systems, with a particular focus on leveraging Convolutional Neural Networks (CNNs) featuring dense architectures. Charles Darwin's groundbreaking work on facial expressions as indicators of human emotional states inspires this research, which aims to elevate FER systems for applications spanning therapy, human-machine interactions, and diverse domains like healthcare, education, and entertainment. While previous studies have recognized CNNs' ability to improve FER accuracy through intricate feature extraction, the evolving nature of the field calls for exploring novel CNN architectures and techniques to further enhance precision and efficiency. In this study, we develop and implement a deep learning model capable of classifying images into seven discrete emotion categories, representing universal human emotions. This objective is achieved by implementing a five-block CNN-based learning algorithm consisting of 44 layers designed to progressively capture complex facial features and expressive patterns. The CNN model performed well on FER and FERG datasets, with accuracy rates of 0.97 and 0.98, showcasing proficiency in facial expression classification. Comparative analysis highlighted its competitive accuracy, emphasizing the importance of feature extraction and architecture design. This research advances facial emotion recognition …

https://scholar.google.com/citations?view_op=view_citation&hl=en&user=rQvaaoMAAAAJ&cstart=20&pagesize=80&citation_for_view=rQvaaoMAAAAJ:Wp0gIr-vW9MC

English