Offline Computer-Synthesized Font Character Recognition Using Machine Learning Approaches

The topic of offline character recognition has long been an interesting issue in the field of pattern recognition. Several experiments have shown that the Neural Network works better in image recognition. The core intention of this paper is to incorporate an efficient and reliable technique for offline computer-generated Latin font recognition. The framework uses several machine learning approaches, such as Support for vector machines (SVM), Random Forest (RF), Recurrent Neural Network (RNN), and Convolutional Neural Network (CNN), like standard classifiers for the recognition. This study compares the performance of these approaches on a benchmark dataset Chars74k with the help of three features extraction techniques. Results show that the deep neural network CNN classifier outperforms the rest without sacrificing performance. The results of these experiments are displayed in various tables and evaluated with the aid of a few evaluation metrics. There are few graphical representations where the rise in the recognition can be confirmed concerning the epoch.