Real time handwriting recognition for mathematic expressions using Hidden Markov Model

• Published in: 2016 International Seminar on Intelligent Technology and Its Applications (ISITIA) pp. 1 - 6
• Main Authors: Pranoto, Yuliana Melita, Setyati, Endang, Pramana, Edwin, Kristian, Yosi, Budiman, Renato
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2016
• Subjects: codebook; codeword; Feature extraction; handwriting; Handwriting recognition; Hidden Markov Model; Hidden Markov models; mathematic expression; Mathematical model; Training; Training data
• DOI: 10.1109/ISITIA.2016.7828623

Mathematic is an important subject, even in our daily live we use mathematic all the time. Calculator as a major tools to help calculate mathematic formulas has become a major requirement in mobile or desktop computer use. Calculator App nowdays can handle basic to complex mathematic formulas. Meanwhile, there are more and more touchscreen-based gadgets nowdays. This simple fact sparks the idea to do a research on online handwriting-based calculator so people can directly write the formula and get the result. This paper is the first phase of research to recognize mathematic expression from user handwriting. Hidden Markov Model (HMM) algorithm is chosen because this is one of the most used algorithms in pattern recognition, such as voice recognition, handwriting recognition, POS tagging and gesture. Every input from handwriting will be processed in several phases, starts from preprocessing and feature extraction. These features will then be transformed into a form of codeword based on codebook which is built by using training data with Vector Quantization. These set of codewords are then compared with HMM models previously built with training data. Experiment was performed covering two things: feature modication experiment and codewords number experiments. Best result is gained for four features combination and 60 units of codewords.