Face Orientation Recognition Based on Multiple Facial Feature Triangles

Abstract

This paper proposes a new method that combining multiple feature triangles with BP neural network, to improve the efficiency and accuracy of face orientation recognition. Based on the traditional indexthe inverted triangle formed by pupils and nasal tip, we find another feature triangle formed by nasal tip and corners of mouth. First we do image preprocessing which includes smoothing linear filter, edge detection and so on. Then both rough and precise detection of feature points are done. Next we extract feature triangle based on two-dimensional plane. Finally BP neural network is used for face orientation recognition. Experimental results show that an approximately 90% success rate is achieved. They also reveal that our new method improves the recognition effect.