A retrospective study on handwritten mathematical symbols and expressions: Classification and recognition

Abstract

Context:Many scientific and technical literature documents contain MSs and MEs that are more challenging to be recognized by computers than plain text. The recognition of HMSE becomes not only an ambitious task but a motivating research area covering concepts of computer vision, pattern recognition, feature extraction, and artificial intelligence. Objective:The objective is to perform an extensive state of the art on the techniques and methods used for recognizing and classifying HMSE. The authors endeavor to bring out all significant findings in sub-processes, representation models, algorithms, tools, datasets, and comparative analysis of the accuracy of the recognition models. Method:The current research implements the standard SLR method based on a comprehensive set of 120 articles published in 21 leading journals and 39 premier conferences and workshops. Results:Existing literature about recognition techniques and models is classified broadly into three categories; AI technique (65%) is majorly implemented in the selected studies. The prominent sub-process ‘segmentation’ (52%) is mostly used. The box and tree are the prevailing representation models. The popular datasets are recognized as CROHME 2014 and CROHME 2016, used by 60% of the selected studies. Masaki Nakagawa, C. Viard Guardin, Richard Zanibbi, and Harold Mouchere are the most noticed authors in ME recognition. Conclusion:The reviewers call for increased awareness of the potential benefits of existing and emerging recognition techniques and identify the need to develop a more accurate and semantic-based recognition model. Recommendations are given for future research.