SMART OBJECT DETECTION AND SPEECH GUIDANCE SYSTEM FOR THE VISUALLY IMPAIRED

Abstract

An intelligent assistive system that provides object detection and voice instructions is proposed to aid the visually impaired in performing daily tasks. The system includes a camera for capturing video feed from which objects are detected using a deep learning technique referred to as YOLO. Spatial positions of objects including left, center, or right and distances between the user and objects are calculated. Detected data is translated into voice instructions using the text-to-speech technology. Stereo sound is employed in order to aid with directionality. The model is trained using data from COCO dataset to provide accuracy in object recognition. The system is relatively light-weight and does not involve the use of expensive equipment thus affordable. It functions in real-time without any delay in response. The system is effective both indoors and outdoors.