The Importance of Pen Motion Pattern Groups for Semi-Automatic Classification of Handwriting into Mental Workload Classes

In this paper, we introduce the pen motion pattern groups (PMPGs) and their contribution to the classification of handwriting into cognitive mental workload classes. We demonstrate the importance of PMPGs by providing an efficient semi-automatic machine learning-based classification framework that distinguishes between handwritten texts written by the same person under different mental workloads. Our evaluation framework is non-language-dependent since we used stroke features, which are not language-specific, and it takes into account the variability in behavioral biometrics between different writers. The handwritten text data was collected using the Computerized Penmanship Evaluation Tool. This digitizer provided accurate temporal measures throughout the writing. As a first stage, the participants were asked to write a given text in the Hebrew language. Then, as a second stage, the participants’ cognitive workload was manipulated by asking them to hold a number in their memory during the entire writing task. In our experiments, we show that incorporating the PMPGs into the classification process yielded an average cognitive load discrimination accuracy of 92.16%, which decreased to 72.90% when the PMPGs were not considered. The separation of handwritten strokes into PMPGs allows us to account for the fact that the strokes are affected differently under different cognitive mental workloads. This novel distinction between PMPGs is important since the handwriting process in each PMPG is different from a perceptual motor and brain-hand control point of view. Moreover, most of the features that are influenced by cognitive workload are those that cannot be discerned by an expert when looking at a handwritten text on paper, such as azimuth, tilt, velocity, acceleration, and pressure.