Segregating Offline and Online Handwriting for Conditional Classification Analysis

Abstract

Offline and online handwriting patterns vary when exposed to external factors. Recent researches on both type handwriting analysis were focused on the image feature extraction, pattern recognition, and classification approach. However, no studies have considered segregating the conditional effects of handwriting patterns on two categories: offline vs. online and normal vs. vibration. Hence, the main goal of the study was to investigate the effects of classifying induced vibration on the offline and online handwriting patterns. There were 25 experimental handwriting samples retrieved under four pre-defined class: FLN, FLV, NLN, and NLV. Extracted data attributes consisted of seven handwriting size metrics and two demographic data. JRip algorithm and Decision Stump algorithm of WEKA tool were employed on tenfold cross-validation mode for handwriting classification and a further segregation by offline and online handwriting. JRip and Decision Stump merely resulted in 54% and 44% classification accuracy respectively. On data categorization between the offline and online handwriting, both algorithms achieved classification accuracies of 56% and 76% respectively. Findings showed that the misclassifications between offline and online were reduced to 5 and 7 instances for JRip and Decision Stump algorithms respectively, whereas the total misclassification between normal and vibration were reduced to 22 and 13 instances.