Support vector machine (SVM) is a recent method to classify the data. SVM has been proved as a powerful tool for solving classification problem. The problem with complex dataset incurs significant complexity while classifying and its efficiency also cost very much. We propose a reduced set support vector machine based on Eigen structure, to classify dataset having multiple features. In this paper, Eigen vectors use to present the whole data in reduced dimensions. This minimize the task of classification by propose method and cost is reduced while efficiency is improved with the increase complexity of data. The proposed method takes a random chunk of data followed by Eigen structure use to reduce the dimension of the data. So as classification problem solve efficiently. We have compared the proposed method with SVM and RSVM. The result signifies that the proposed method gives better result in comparison to SVM and RSVM.

Wavelet Transform is basically used for magnitude depletion. It is used for axing the proportion of picture. Including good multi-resolution and multi-scale analysis, wavelet transform also has the propensity of denoting local signal attribute by using the high and low pass filtering, image can be decomposed into divergent scales of approximation components. But in wavelet transform, the higher decomposition layers will lost a lot of information, by which reduce the recognition rate. 2DPCA is a sort of image extraction method deal directly with image data and does not need dimension reduction. It is undertaking image data without step of vectorization. However 2DPCA algorithm reduces the computational complexity, it takes up more storage space. In that paper proposed an advanced technique which is stationed on dual-tree complex wavelet transform and improved 2DPCA employing SVM classifier in order to give higher coherent recognition rate. The experimental results on ORL and YALE face databases shows that the proposed method improves the performance of face recognition with respect to exiting techniques.