Face Classifier Based on Boosting

Abstract

We have proposed a new classifier which is able to handle multi-class discrimination problem very easily. The classifier consists of Boosting (AdaBoost) units and Fuzzy Partition(FP) unit. The architectre of Boosting + FPM leads to not only promotion of classification precision but also real-time classification without degrading precision. We have applied Boosting + FPM to classify face images to four class (puerility, youth, middle age, advanced age) As a result, Boosting + FPM gives more than 79 and FPM without Boosting gives 63% and 48% respectively. In addition to that, Boosting + FPM gives real-time classification which precision is not less than 72%.