A surveillance robot with human recognition based on video and audio

Abstract

Household security becomes more and more important with the growth of aged population. A surveillance robot is a potential solution for this issue. In this paper, we present a novel household surveillance robot with human recognition. Since a master-slaver structure and efficient interfaces are used for its hardware, this robot can always make prompt response to the outside information, even though it has integrated a lot of sensors such as audio, video, gas sensors, and moving controllers. To locate the abnormal sound, we use the TDOA (Time Difference of Arrival) method. And in our approach, the haar-like feature and the weighted LBP (Local Binary Patterns) algorithm are employed to detect and recognize faces. When abnormal sound occurs, the robot locates it and turns to the direction and make sure whether a friend or an intruder comes. After testing the rate of success of abnormal events detecting and people recognition, it shows that the robot has a good performance in both our experiments and real environments.