Development of a Cost-Effective, AI-Powered Assistive Wearable for the Visually Impaired Using OTG Endoscope Technology and Smartphone Edge Computing

Visual impairment significantly restricts independent mobility and environmental awareness, especially in developing nations where advanced assistive technologies remain prohibitively expensive. Existing smart-wearable prototypes frequently rely on standalone microcontroller platforms such as the ESP32-CAM, which suffer from severe thermal throttling, limited memory, and bulky battery requirements when deployed for continuous real-time computer vision tasks. This paper presents Vision-Link AI, a novel, cost-effective wearable assistive device that circumvents these hardware limitations through a hybrid architecture. The system utilizes a lightweight OTG (On-The-Go) endoscope camera integrated with a standard smartphone via edge computing for visual perception, coupled with a custom-built, frame-mounted Bluetooth audio receiver for discreet spatial feedback. By offloading the visual computational workload to the smartphone's processor using optimized, quantized deep-learning models, the system achieves real-time obstacle detection with an end-to-end response latency below 120 ms. The integrated hardware, comprising a custom power-distribution circuit and a class-D audio amplification chain driven by a 3.7 V lithium battery dedicated solely to audio and illumination, delivers instant Text-to-Speech (TTS) guidance while remaining cool to the touch under continuous operation (approximately 27 °C ambient). This hybrid architecture drastically reduces cost relative to both cloud-dependent and fully embedded designs, prevents the thermal issues characteristic of ESP32-class wearables, and provides a highly scalable, ergonomic solution for the visually impaired.