Publication: Exploration of handwritten allographic features attributed to the primary educational backgrounds of Malaysian writers for forensic comparison
Loading...
Date
2023-09
Authors
Gannetion, Linthini A/P
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Class characteristics in handwritings is useful in writer profiling when
exemplar samples are unavailable for comparison. The unique education system in
Malaysia may have established specific characteristic handwritten allographic features
that deserve investigation within the intelligence context. This study aims to
investigate the common and distinguishable handwriting allographic features
attributed to the primary education backgrounds of Malaysian writers for forensic
comparison. At the preliminary stage, 50 sets of handwriting samples regardless of
primary education background were thoroughly reviewed to determine the general
handwritten features. Subsequently, characteristic handwritten features potentially
attributed to the primary education backgrounds (i.e., National Schools, Chinesemedium
vernacular schools, Tamil-medium vernacular schools, and Islamic Religious
Schools) were identified and their associations among the four backgrounds were
analysed using Chi-square test followed by multinomial logistic regression (MLR).
Selected handwritten features were then examined on newly recruited 400 writers (100
writers for each background) in the field study. A classification and regression tree
(CART) was established, and the predictive capability was determined using blind
exemplars. From a total of 189 features determined at the preliminary stage, 23 of them
were significantly attributable to primary education backgrounds (p-value <0.05)
based on the Chi-square test and MLR model. In the field study, the number of
statistically significant features were further reduced to 17 features, comprising of letters ‘B’, ‘D’, ‘E’, ‘G’, ‘H’, ‘I’, ‘p’, and ‘t’, and nine features were recognised to be
capable of distinguishing writers from a single education background from the others.
Through CART, five handwritten features carrying greater importance in
discriminating the writers were determined, including letter ‘t’ with crossbar written
first followed by vertical stroke, letter ‘B’ with no protruding initial stroke, letter ‘E’
constructed with three individual strokes, letter ‘E’ with the stem written first followed
by the top arm, crossbar and base arm, and letter ‘G’ with stress written first followed
by chin and barb. It allowed for a 71.0% accurate classification, and testing on blind
exemplars demonstrating an 85.0% prediction accuracy. To conclude, this study had
successfully determined the handwritten allographic features attributed to the primary
educational backgrounds of Malaysian writers, providing the foundation data for the
establishment of forensic intelligence framework for handwriting examination, which
would be beneficial to the investigation of document related cases, especially when
reference samples are not available.