Multi-angle face expression recognition based on integration of lightweight deep network and key point feature positioning

Abstract

In the current advancement of science and technology, robot and machine vision technology have significant application value. Due to their mimetic structure, quadruped robots are better equipped to adapt to the requirements of walking over complex terrain. This paper introduce lightweight deep network and combining key point feature positioning for multi-angle face expression recognition. Using robot dog to recognize facial expressions will be affected by distance and angle. To solve this problem, this paper proposes a method for facial expression recognition at different distances and angles,which solved the larger distance and deflection angle of face expression recognition accuracy and real-time issues. The dataset used in this study was collected by adjusting the camera mounted in an AI robotic dog in a laboratory scenario. The result shows a common and fast and intuitive visual representation of human beings to convey emotions or mental states, facial expression is an effective way for robots to obtain information and interact with people through visual images.