Face orientations recognition is the premise of face recognition. Many approaches have been proposed to recognize the human face orientations, but these methods usually rely on the assumption of a frontal view of human face. Besides this, the costs of different types of misclassifications are treated equally, without considering the number of orientations. This may lead to the inefficiency of face recognition system. Aiming at these problems, this paper aims to design a convenient classification system for recognizing face orientations. Firstly, the classification model is established based on PSO-SVM with a hybrid kernel. Secondly, maturity metric (M) is structured with precision metric (P) and cost metric (E) according to the classification results. It is used to judge whether the classification model meets the performance requirement. The experimental results show that the proposed classification system produces P=0.9815 and E=0.011, and the value of M equals 83.33. It indicates that the proposed classification system can be used for face orientations recognition in advance, which will contribute to the success of face recognition.