Sensor Enabled Proximity Detection with Hybridisation of IoT and Computer Vision Models to Assist the Visually Impaired

Abstract

Proximity Detection Systems (PDS) are used to detect objects or persons close to Visually Impaired (VI) persons. Sensors are used to identify proximity based on the distance from objects. This study aimed to design a hybrid proximity detection framework for VI people using ultrasonic sensors embedded in a Raspberry Pi board to detect the proximity of a VI user in an office environment. Hybridization was based on the integration of IoT-enabled devices, ultrasonic proximity sensors, and computer vision algorithms to control the detection of objects or people and inform the user with a voice message. The model framework was implemented with 100 samples and tested with 10 analyses in each sample. The results showed significant improvement in detecting the proximity of the objects with an accuracy of 98.7%, outperforming current PDS with good results in precision, range, obstacle recognition, false positives and negatives, response time, usability, durability, reliability, etc.