Human ear recognition is one of the most important branches in biometrical recognition and artificial intelligence fields. In this paper, considering the unique texture feature of human ear image, the spatial pyramid visual bag-of-words model was adopted. It transforms the relatively low-level local descriptors of human ear images into global features to preserve the high-level semantic meanings. The support vector machine classifier is utilized to perform the training and recognition task. Experimental results demonstrate that the adopted model could achieve a better accuracy, as an extension and exploration in human ear recognition methods.