Mining Vibrational Effects on Off-line Handwriting Recognition

Authors

  • L.C. Wong School of Mechanical Engineering, Engineering Campus, Universiti Sains Malaysia, 14300 Nibong Tebal, Penang, Malaysia.
  • W.P. Loh School of Mechanical Engineering, Engineering Campus, Universiti Sains Malaysia, 14300 Nibong Tebal, Penang, Malaysia.

Keywords:

Classification, Handwriting Recognition, Offline Handwriting, Parkinson’s Disease, Vibrational Stress,

Abstract

An individual’s handwriting exhibits variation under external factors, such as writing surface, writing pen, and writing force. Recent studies on handwriting recognition emphasised on interpretation techniques using feature extraction, pattern recognition, and classification approaches. However, no study has evaluated the effects of external source vibrations on handwriting patterns. Hence, this study analyses offline handwritings features on two conditions: with vibrational (V) and without vibrational (N) stresses using the data mining approach. The goal was mainly to recognise individual handwriting features characterised by vibrational conditions. This research was performed on experimental and public offline handwriting databases consisting of English phrases written under (V) and (N) conditions. Vibrational stresses impact was simulated with Mondial Slim Beauty Fitness Massager strapped onto the writing table and Parkinson’s Disease (PD) patient with hand tremor symptom. Nine handwriting size metrics with demographic data were extracted as the data attributes. PART and J48 classification algorithms in Waikato Environment for Knowledge Analysis (WEKA) tool were employed on cross-validation and full training set modes to classify the handwriting data into two predefined classes: (V) and (N). Further significant attributes that distinguish data classes were examined on the decision list and tree diagram constructed from PART and J48. Findings showed that size of “short” letter and “tail” letter were dominant to determine handwriting classes at accuracies: 55.3%- 66.7% (crossvalidation) and 86.0% - 100.0% (training set). The study suggests that the size of “short” letter and “tail” letter are the dominant features to distinguish between the (V) and (N) handwriting.

Downloads

Download data is not yet available.

Downloads

Published

2018-09-26

How to Cite

Wong, L., & Loh, W. (2018). Mining Vibrational Effects on Off-line Handwriting Recognition. Journal of Telecommunication, Electronic and Computer Engineering (JTEC), 10(3-2), 71–75. Retrieved from https://jtec.utem.edu.my/jtec/article/view/4714