In this paper, the authors propose a new method to construct feature vectors of human iris based on circular technique. Since an iris has a circular shape, the circular construction of the iris feature vector is expected to have higher ability to capture iris' characteristics. Different from the conventional method which constructs iris feature vectors through left to right scanning process, the circular technique scans the pixel-intensity value of iris circularly. As the result, the length of the constructed feature vectors is dependent on the length of iris' radius. In the experiments, various recognition methods of feature vectors were compared. The iris recognition results show that the proposed system with Statistical Euclidean Distance method gives the highest recognition accuracy with 84.5%.