Feature curve metric for image classification

Abstract

In this paper, an improved classifier based on nearest feature line (NFL), shortest feature line segment (SFLS) and nearest feature center (NFC), called the feature cure metric (FCM), is proposed for hand gesture recognition and face recognition. Borrowing the concept from the NFC and SFLS classifiers, the proposed classifier uses the novel distance metric between the test sample and the pair of prototype samples. A large number of experiments on Jochen Triesch Static Hand Posture (JTSHP) Database, Yale face database and JAFFE face database are used to evaluate the proposed algorithm. The experimental results demonstrate that the proposed approach achieves better recognition rate than the other well-known classifiers, such as nearest neighbor (NN) classifier, NFL classifier, nearest neighbor line (NNL) classifier, centerbased nearest neighbor (CNN) classifier, extended nearest feature line (ENFL) classifier, SFLS classifier and NFC classifier.