Bird species recognition is a challenging problem due to the variant illumination and different view point of camera. In this paper, a new feature which is the ratio between the distance of the eye to the root of beak and the distance of the width of the beak is used to distinguish the different bird species. Integrated the new feature into the multi-scale decision tree and the SVM framework, a new bird species recognition algorithm is proposed to get the final recognition result. The Experiment results show that the proposed new feature can improve the correct classification rate about nine percent.