Deep learning for multi-factor human behaviour analysis

Abstract

This research aims to develop a comprehensive, multi-scale human behavior analysis system using state-of-the-art techniques and methodologies. Various approaches were employed, including facial expression recognition, emotion expression classification, and Natural Language Processing (NLP), to analyze human behavior in digital communication contexts. The study utilized benchmark datasets, such as Fer2013, Yale dataset, and BookClub artistic makeup dataset, for training and testing purposes. The proposed system combined engineered features algorithms, metalearning approaches, and ensemble techniques to improve classification accuracy. Novel contributions of this work include the utilization of Homogeneous Underlying CNN Ensembles, a Supervisor ANN to enhance accuracy, and the introduction of a meta-learning approach for facial recognition and emotion expression classification systems under real-life scenarios. The potential of the proposed system to accurately detect and classify human behavior in images, videos, and textual data was demonstrated by this research’s findings. The research’s importance lies in its contribution to the understanding of human behavior in various contexts, with potential applications in fields such as psychology, marketing, and social research. By incorporating advanced methodologies and techniques, this work paves the way for further advancements in the field of human behavior analysis.