Receptionist and Security Robot Using Face Recognition with Standardized Data Collecting Method

Abstract

Face recognition has become the front runner for deep learning applications in the real world and this paper focuses on its implementation in a human-robot interaction and security system. For this specific project, it is inherent that restraints are created to allow the system to produce greater performance within the requirements of a receptionist and security robot. A k-nearest neighbors classifier is applied to further enhance the accuracy of face recognition. By sequencing images from videos, we create large datasets to train our own classifier in various conditions to increase its accuracy and lower false-positive rates in poor lighting environments. With the goal of creating a service robot, we have standardized our method of data collection for new inputs that will assist the recognition process in variable conditions of operation. The resulting product is a system that can accurately predict known and unknown faces with Asian features.