Recent Approaches in Handwritten Word Recognition

The recognition of handwritten characters and words are popular in research fields. Due tounconstrained writing styles along with connected and overlapping characters, handwriting recognitionremains a challenging task. The objective of handwriting recognition is to transform a character, word, ortext into a habitual representation that can be processed by a machine. A typical word recognition systemgoes through different stages such as preprocessing, feature extraction and classification. This systemrequires a word dataset which is collected directly from the people or generated synthetically. Differentapproaches to each stage are discussed in the article. These approaches include sub-network of CNNmodel with low cost trainable parameters, and lexicon-free object recognition model. An evaluationmetrics consisting of word error rate, accuracy, F1-score, precision, and recall are used to measureperformance of the models.