Face recognition based on real adaboost and Kalman forecast

Abstract

In this paper, a novel face recognition method based on Real AdaBoost algorithm and Kalman Forecast is implemented. Real AdaBoost algorithm can obtain great accuracy with machine learning. Meanwhile, Kalman Forecast is introduced to track human faces detected, making face detection more efficient. We tested our new method with many video sequences. The detection accuracy is 98. 57%, and the average processing time on a windows XP, PIV 2.4GHz PC was less than 20 ms for each 640*480-pixel image. So the proposed face recognition method is real-time. © 2011 Springer-Verlag.