Towards a grid based sensor fusion for visually impaired navigation using sonar and vision measurements

Abstract

This work presents an integrated approach on daily mobile devices, sensors, and resources available by Cloud computing providers to improve the navigation of visually impaired people. Input taken from the heterogeneous sensors such as sonar, vision, orientation and inertial sensors are used for object detection and recognition. The proposed approach entails grid based obstacle localization using sonar sensors and precise identification of that obstacle by vision sensors. This is coarse to fine obstacle identification process which increases the accuracy and decreases computational processing overhead. This improves the walking experience of the visually impaired by making them more independent. This framework facilitates both audio and tactile feedback on the obstacles in the close proximity of the visually impaired people. Improvement of the independent navigation of the visually impaired people by using sensors fusion is the key contribution of this research.