In a multi-lingual country like India, identification of the multi-script in an image of a document page is of primary importance for a system processing multi-lingual document. In this paper we present a technique for classify printed Bangla document. Our script classification approach is called SVM based classification. In the classification stage which is the second step of the Optical Character Recognition Engine‥ It includes making decisions regarding membership of class of a pattern which is being studied. The aim is to design a decision algorithm that easily computes and minimizes the probability of misclassification. SVM classification system has been tested on printed Bangla Language script and an average accuracy of 94.78% has been achieved.