Segmenting Handwritten Math Symbols Using AdaBoost and Multi-scale Shape Context Features

Abstract

Abstract-This paper presents a new symbol segmentation method based on AdaBoost with confidence weighted predictions for online handwritten mathematical expressions. The handwritten mathematical expression is preprocessed and rendered to an image. Then for each stroke, we compute three kinds of shape context features (stroke pair, local neighborhood and global shape contexts) with different scales, 21 stroke pair geometric features and symbol classification scores for the current stroke and stroke pair. The stroke pair shape context features covers the current stroke and the following stroke in time series. The local neighborhood shape context features includes the current stroke and its three nearest neighbor strokes in distance while the global shape context features covers the expression. Principal component analysis (PCA) is used for dimensionality reduction. We use AdaBoost with confidence weighted predictions for classification. The method does not use any language model. To our best knowledge, there is no previous work which uses shape context features for symbol segmentation. Experiment results show the new symbol segmentation method achieves good recall and precision on the CROHME 2012 dataset.

Offline Features for Classifying Handwritten Math Symbols with Recurrent Neural Networks

“…In addition to the classes detailed in TOP 1 computation, many errors were caused by very similar classes like: {5, s}, {t, +}, {comma, )}, {q, 9} or {z, 2}. There are not published results yet using this database, but an experiment using the described online features and BLSTM-RNNs is reported in [6] on a smaller dataset such that it outperformed previous publications [4], [18].…”

“…We report top-1, top-2 and top-5 recognition rates. Furthermore, there are several classes that produce many classification errors because they have very similar shape but different semantic [18], [6]. Hence, we also computed top-1 recognition rate where those similar classes were merged (TOP 1 ).…”

“…Furthermore, we also wanted to assess the performance of a set of features based on polar descriptors because, in the last years, several approaches based on polar shape descriptors [14] have been published in the mathematical expression recognition field. For example, polar features have been used to detect layout classes for symbols [17], symbol segmentation [18] or spatial relationships classification [19]. There are also proposal for handwritten text recognition using polar histograms [15].…”

“…Since the MathBrush database [20] was released more papers are reporting comparable results [4], [18], [6]. These papers presented online math symbol recognition experiments.…”

On-line Handwritten Mathematical Expression Recognition Method Based on Statistical and Semantic Analysis

“…Ha et al [1] proposed a recursive X-Y cut segmentation method, which worked well on typeset MEs but not fitted for handwritten cases. Zanibbi et al proposed an AdaBoost and PCA method in [2] but lacked a language model. As for symbol recognition module, there are numerous researchers focusing on on-line handwritten recognition and mature methods have been proposed [4].…”