Ear recognition method based on fusion features of global and local features

In the paper, we propose a new method for Ear recognition. Firstly, we extract global features using Kernel Principal Component Analysis (KPCA) technique and extract local features using Independent Component Analysis (ICA) technique. Then we establish a correlation criterion function between two groups of feature vectors, extract their canonical correlation features according to this criterion, and finally form effective discriminant vectors for recognition. For validation of our method, we have tested our method on the USTB ear database by using linear Support Vector Machine. Meanwhile, we have compared performance of our method with that of KPCA-based and ICA-based methods. The experiment results show the performance of our method is superior to those of other methods.