A SVM-HMM Based Online Classifier for Handwritten Chemical Symbols

Abstract

This paper presents a novel double-stage classifier for handwritten chemical symbols recognition task. The first stage is rough classification, SVM method is used to distinguish non-ring structure (NRS) and organic ring structure (ORS) symbols, while HMM method is used for fine recognition at second stage. A point-sequence-reordering algorithm is proposed to improve the recognition accuracy of ORS symbols. Our test data set contains 101 chemical symbols, 9090 training samples and 3232 test samples. Finally, we obtained top-1 accuracy of 93.10% and top-3 accuracy of 98.08% based on the test data set.

Chemical Symbol Feature Set for Handwritten Chemical Symbol Recognition

“…For trajectory-based symbol recognition approaches, different types of chemical features have also been extracted. In [8,7], Zhang et al proposed 11 types of chemical features including normalized stroke point coordinates, normalized first derivatives and second derivatives of the stroke points, curvature, writing direction, aspect, curliness and linearity for symbol recognition in HMM and SVM-HMM. In [6], Tang et al used some simple chemical features including number of strokes, stroke point coordinates, horizontal angle and turning angle for the SVM-EM symbol recognition.…”

“…For the past decade, different techniques such as Support Vector Machines (SVM) [3], Hidden Markov Models (HMM) [8], hybrid SVM-HMM [7] and Support Vector Machine-Elastic Matching (SVM-EM) [6] have been proposed for chemical symbol recognition. The current image-based approaches [3,5,4] mainly considered the geometrical and statistical information from the captured images of users' handwriting strokes.…”

“…As such, users' writing styles during the writing process are not captured. On the other hand, the current trajectory-based recognition approaches [8,7] only considered extracting temporal symbol features based on stroke points. These techniques only focused on capturing features about the writing process which are dependent on users' writing styles.…”

Online stroke segmentation by quick penalty‐based dynamic programming

“…Sketching on paper or a pen‐input display such as table PCs and smart phones is a natural way to quickly externalise ideas, and it is often used in the early prototyping stages to express and record design ideas and can be adopted by many domains including mechanical engineering [1], software design [2, 3], information architecture [4, 5] and pen‐based recognition and simulation systems [6]. As a recent trend against traditional window, icon, menu and pointer paradigm, sketching plays a growing important role in two‐dimensional (2D)/three‐dimensional (3D) modelling and human–computer interface [7].…”

Online structural analysis for handwritten chemical expression recognition

“…Symbol recognition [5] converts user input stroke data into the corresponding chemical symbols, whereas structural analysis analyzes the sequence of recognized chemical symbols with structural information and converts them into the corresponding chemical expression representations. In this paper, we propose an effective and novel approach to perform structural analysis on chemical expressions.…”