Human Face Detection Enabled Smart Stick for Visually Impaired People

Abstract

The present work enhances the capabilities of a newly developed smart stick (Sharma et al Multiple distance sensors based smart stick for visually impaired persons, Las Vegas, pp 1–5, 2017 [1]) by detecting human faces using the PI camera on Raspberry Pi board. Visually impaired people can use this stick developed by us (Sharma et al Multiple distance sensors based smart stick for visually impaired persons, Las Vegas, pp 1–5, 2017 [1]) to locate static and dynamic obstacles using multiple distance sensors and now can even detect the presence of a human if he/she is in front of the user. The problem of human face detection with simple and complex backgrounds is addressed in this paper using Haar-cascade classifier. Haar classifier has been chosen because it does not require high computational cost while maintaining accuracy in detecting single as well as multiple faces. Experimental results have been performed on the smart stick in indoor and outdoor unstructured environments. The stick is successfully detecting the human face(s) and generates alerts in form of vibration in the stick as well as audio in a headphone. OpenCV-python is used to implement Haar-cascade classifier and an accuracy ≈98% is achieved with this setup.