Hybrid CNN-SVM Classifier for Handwritten Digit Recognition

• DOI: 10.1016/j.procs.2020.03.309
• Publication Date: 2020-04-16T15:52:15Z
• Authors: Savita Ahlawat; Amit Choudhary
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Abstract The aim of this paper is to develop a hybrid model of a powerful Convolutional Neural Networks (CNN) and Support Vector Machine (SVM) for recognition of handwritten digit from MNIST dataset. The proposed hybrid model combines the key properties of both the classifiers. In the proposed hybrid model, CNN works as an automatic feature extractor and SVM works as a binary classifier. The MNIST dataset of handwritten digits is used for training and testing the algorithm adopted in the proposed model. The MNIST dataset consists of handwritten digits images which are diverse and highly distorted. The receptive field of CNN helps in automatically extracting the most distinguishable features from these handwritten digits. The experimental results demonstrate the effectiveness of the proposed framework by achieving a recognition accuracy of 99.28% over MNIST handwritten digits dataset.