Edge-IoT and MLLMs for Education and Scene Understanding: Assisting Vision and Hearing-Impaired Individuals

The rapid advancement of IoT and edge computing technologies has opened new horizons for creating assistive solutions tailored to individuals with sensory impairments, particularly those with hearing and vision disabilities. This article presents a novel edge-IoT-based framework that integrates multimodal large language models (MLLMs), multi-object tracking, and scene understanding to develop real-time, responsive assistance for impaired individuals. Our proposed system is designed to enhance accessibility and quality of life by providing educational tools and entertainment options that cater specifically to the needs of this community. The proposed solution leverages the computational power of edge devices to process data locally, ensuring low latency and high responsiveness, which are critical for real-time applications. Furthermore, we explore the potential of generative AI models in improving autonomy, with a particular focus on real-time transcription services for the hearing impaired and scene description services for the visually impaired. This work demonstrates the feasibility and effectiveness of using edge-IoT technologies combined with advanced AI techniques to create inclusive environments that empower disabled individuals, ensuring that technological advancements are leveraged to foster accessibility and independence.