Dynamic object recognition and command detection for assisting blind people using image processing techniques

Abstract

Human senses are the only medium to communicate with environment. Our brain makes a combination of different senses and convert it into a meaningful mixture. We, humans have five different senses as in two eyes to see, one nose to smell, two ears to hear, one tongue to taste and a skin to feel touch. All of these senses are connected deeply with each other. One sense always has the backup of another one. There are thousands of people in the whole world who are blind. They face many kinds of hindrances in their everyday life. They suffer badly in travelling here and there like shopping, work, educational institutions and so on due to their problem in sight. Hence, many papers have been published on making blind or visually impaired people’s navigation easier. Different papers proposed different methods and different ways to navigate their way. Our paper aims to represent a proposed system that helps a visually impaired, which detects various objects and finds right path for them to reach to their destination. Our projected model is mainly based on a well-known object detect and identifier algorithm that needs to be trained by a dataset. Step by step, simulations using dynamic object detection algorithm results in a better accuracy to detect those objects in our proposed model. Hereby, our paper presents the idea of an object detection system based on object extractions; networks for segmentations match with the recognized dataset and locate those objects from the live video-image extractions, which gives an audio output.