A new ensemble based classifier using feature transformation for hand recognition

This paper presents a new method for user identification based on hand images. Because in user identification the processing time is an important issue, we use only the hand boundary as a hand representation. Using this representation, an alignment technique is used to make the index of corresponding features of different hands get the same index in the feature vector. To improve the classification performance a new ensemble-based method is proposed. This method uses feature transformation to create the needed diversity between base classifiers. In other words, first different sets of features are created by transforming the original features into new spaces where the samples are well separated, and then each base classifier is trained on one of these newly created features sets. The proposed method for constructing an ensemble of classifiers is a general method which may be used in any classification problem. The results of experiments performed to assess the presented method and compare its performance with other alternative classification methods are encouraging.