REAL-TIME FACE RECOGNITION SYSTEM AS PERSONAL ASSISTANT FOR PEOPLE WITH BLINDNESS

Recognizing others is a significant challenge for people with blindness and can hinder their social activities. This work’s primary goal is to build a real-time face recognition system to support people with blindness in recognizing other people around them. The system is built on an Android platform and functioned as a personal assis- tant. It implements Haar-cascade classifier and local binary pattern histogram methods to detect and recognize faces from input video and produces voice as the output. Three scenarios are conducted to measure system performance. Each procedure is repeated three times and experimented with different distance values between 1 and 2.5 meters and vari- ous threshold (T) values between 45 and 90. Besides, this study also involved a parameter grid for system performance evaluation. The test scenario shows the best performance is shown in scenario 1 with the user did not move. Hence, the camera’s position remained steady and could capture stable images. The experiment shows that the system results in the highest average accuracy of 90.77%, using a threshold value of 90 and 8 × 8 for grid parameters.