Gender classification based on multi-classifiers fusion for Human-Robot interaction

In the robotic area, robot will take some actions depending on gender or person. For example, according to gender, robot can change different listening-modes for voice recognition to improve recognition accuracy. Besides, we can recognize face based on gender to change predicting model from male or female database. Therefore, we develop a Human-Robot interaction through gender recognition. In this paper, we adopt multiple classifiers based on support vector machine to recognize gender in low-resolution facial images (36-by-36 pixels); because fusing multiple classifiers usually promises higher classification accuracy than using individual classifier. Therefore, we conduct our research on comparing bootstrap aggregating (Bagging) and Adaboost. In conclusion, we find Adaboost with image pixels as input indeed facilitates the gender classification.