Abstract: A method of ear recognition is presented.Firstly,the original ear image is preprocessed.Secondly,a method of edge detection based on gray-scale morphology and local threshold segmenting is used to extract edge information of the auricle contour.Thirdly,a new modified Hausdorff distance based on the standard deviation and length difference of edge lines is adopted for ear feature description.Finally,the support vector machines are trained to perform ear recognition.Experimental results show that our proposed method can achieve a higher ear recognition rate.