Enhancing Social Interaction for the Visually Impaired: A Systematic Review of Real-Time Emotion Recognition Using Smart Glasses and Deep Learning

This systematic review investigates the application of smart glasses as an assistive technology for visually impaired individuals, focusing on real-time emotion recognition powered by deep learning (DL) models. This study analyzed the advancements in hardware (e.g., Raspberry Pi and NVIDIA Jetson) and software (e.g., CNN and RNN) that enhance facial expression detection. This review identifies the integration of cameras, processors, and audio outputs as being pivotal for assistive devices. Furthermore, it discusses the technical challenges that affect the reliability of emotion recognition, such as environmental variability, cultural diversity in emotional expressions, and scalability for real-world use. Despite these challenges, this review highlights the transformative potential of combining DL models with wearable technologies to improve the quality of life of visually impaired individuals. These findings suggest that improvements in DL models and hardware design can significantly enhance user independence and social interaction among visually impaired individuals using smart glasses. Moreover, this study underscores the need for multimodal systems that incorporate voice, facial, and environmental data to overcome the existing limitations and adapt to individual user needs. This survey builds on an analysis and covers various aspects of enhancing social interaction for the visually impaired, which remain underexplored.