AI-driven emotion recognition systems for sustainable mental health care: an engineering perspective

Abstract

Emotion recognition systems are transforming human-computer interaction (HCI) applications by enabling AI-driven, adaptive, and responsive mental health interventions. This study explores AI-based emotion recognition technologies using facial expressions, voice analysis, text-based sentiment processing, and physiological signals to develop scalable, real-time mental health support systems. Utilizing datasets such as FER2013, JAFFE, and CK+, our research examines deep learning models, including EfficientNet XGBoost, which achieved over 90% accuracy across key evaluation metrics. Unlike traditional mental health interventions, AI-driven systems provide cost-effective, accessible, and sustainable solutions through telemedicine, wearable biosensors, and virtual counselors. The study also highlights critical challenges such as algorithmic bias, ethical AI compliance, and the energy consumption of deep learning models. By integrating machine learning, cloud-based deployment, and edge computing, this research contributes to the development of sustainable, ethical, and user-centric AI solutions for mental health care. Future directions include AI model optimization for energy-efficient deployments and the creation of diverse, inclusive datasets to improve performance across global populations.