Optical Character Recognition Using Minimal Complexity Machine and Its Comparison with Existing Classifiers

摘要

Optical character recognition is an important field of research with widespread applications. Researchers have endeavored to replicate the human ability to read printed text, and extensive research has been carried out. In this paper, we recognize optical characters using a recently proposed classifier minimum complexity machine (MCM). To aid in performance analysis, with existing classifiers we have compared the results obtained from MCM with results from support vector machine (SVM) and k-nearest neighbor (k-NN). A common dataset for testing and training was used throughout to determine the accuracy and time taken for testing. Principal component analysis has been used to obtain a reduced set of features prior to the training and testing process. An analysis of the effect of the number of components on the accuracy of minimum complexity machine, support vector machine, and k-nearest neighbor has been provided as well. It was noted that minimum complexity machine has given accuracy comparable to that of the existing classifier though the time taken for testing was substantially reduced.