Eye detection using discriminatory Haar features and a new efficient SVM

This paper presents an accurate and efficient eye detection method using the discriminatory Haar features (DHFs) and a new efficient support vector machine (eSVM). The DHFs are extracted by applying a discriminating feature extraction (DFE) method to the 2D Haar wavelet transform. The DFE method is capable of extracting multiple discriminatory features for two-class problems based on two novel measure vectors and a new criterion in the whitened principal component analysis (PCA) space. The eSVM significantly improves the computational efficiency upon the conventional SVM for eye detection without sacrificing the generalization performance. Experiments on the Face Recognition Grand Challenge (FRGC) database and the BioID face database show that (i) the DHFs exhibit promising classification capability for eye detection problem; (ii) the eSVM runs much faster than the conventional SVM; and (iii) the proposed eye detection method achieves near real-time eye detection speed and better eye detection performance than some state-of-the-art eye detection methods.