Handwritten Digit Recognition by Combining SVM Classifiers
Dejan Gorgevik, Dusan Cakmakov
Abstract: Recent results in pattern recognition have shown that SVM (Support Vector Machine) classifiers often have superior recognition rates in comparison to other classification methods. In this paper, a cooperation of four SVM classifiers for handwritten digit recognition, each using different feature set is examined. We investigate the advantages and weaknesses of various cooperation schemes based on classifier decision fusion using statistical reasoning. The obtained results show that it is difficult to exceed the recognition rate of a single, well-tuned SVM classifier applied straightforwardly on all feature sets. In our experiments only one of the cooperation schemes exceeds the recognition rate of a single SVM classifier. However, the classifier cooperation reduces the classifier complexity and need for training samples, decreases classifier training time and sometimes improves the classifier performance.