Handwritten Character Recognition Using Unsupervised Feature Selection and Multi Support Vector Machine Classifier
International journal of intelligent engineering and systems · 2021
In recent times, identifying Kannada, Arabic and English handwritten characters is a challenging task in pattern recognition application.The low resolution, complex backgrounds, text orientation, text size, and the variations in the writing styles makes character recognition as a challenging task.To address the above stated issues, a new automated character recognition model is introduced in this paper.Firstly, skewed line segmentation technique is applied to the handwritten character for dissecting the document images into line elements and then split into phrases and single characters.Next, AlexNet model is used for extracting the deep features from the individual characters, where the extracted feature vectors are multi-dimensional in nature that increases the system complexity.So, an unsupervised feature selection algorithm is proposed to select the active feature vectors that are fed to multi support vector machine classifier for individual character classification such as 64 classes in English language, 10 classes in Arabic language, and 657 classes in Kannada language.Experimental analysis showed that the proposed model obtained 85.80%, and 95.55% of accuracy on the chars74K dataset for Kannada, and English characters, respectively.In addition, the proposed model obtained 71.79%, and 99.97% of recognition accuracy in Kannada and Arabic handwritten character recognition on a real time dataset and MADbase digits dataset.The obtained results are better compared to the existing deep learning models; DIGI-Net, feed forward neural network, and context aware model.