This paper presents a robot-in-the-loop face detection and recognition system. This system is constructed based on a robot coordinaive control platform. By means of the coordinaive control platform, robot can actively capture high quality face images and move according to the image acquisition request, so that the performance of both the face detection and recognition can be remarkably improved. To improve the speed of face detection, a hybrid method consisted of Adaboost algorithm and Skin Color Model for face detection is proposed. In addition, the Embedded Hidden Markov Model (EHMM) is employed to recognize the detected faces. We demonstrate the effectiveness of he proposed approaches on several datasets. The superior property of the proposed system is also shown on a real robot system in the uncontrolled indoor conditions. © 2013 IEEE.