Handwritten character recognition using optimization based skewed line segmentation method and multi-class support vector machine

Handwritten character recognition (HCR) has become a growing research, owing to its several applications in processing the images, recognizing the patterns, communication technologies, etc. However, HCR is affected because of different styles of writers, or even if one writer has a different type of writing style according to the situation. Therefore, it is a tedious task to efficiently extract and recognize the digits or characters in the document/image. In that, HCR and classification is the toughest task in the field of pattern recognizing due to the various writing instruments and styles obtained from dissimilar widths, orientations, and sizes. In this work, multi-class support vector machine (MSVM) classifier is utilized for character identification. At first, the handwritten images are developed from real-time and Chars74k datasets, and then, pre-processing is executed for character image enhancement using Binarization. Moreover, the discrete lines are segmented by applying the modified whale optimization algorithm (MWOA)-based Otsu thresholding technique. The proposed MWOA-MSVM gained better evaluation outcomes at 99.2% accuracy, 98.43% precision, 98.99% recall and 97.54% F1-score in HCR when compared to the other techniques that were, hybrid feature-based long short-term memory (LSTM) and stacked sparse auto encoder.