Abstract
This study investigates the feasibility of identifying users in collaborative digital environments by analyzing physical sketching mannerisms. By capturing high-fidelity pen pressure and tilt data, researchers developed a biometric signature for individual users. Results indicate that pen tilt is a remarkably consistent and unique identifier, allowing a linear classifier to distinguish between users with up to 94% accuracy on a stroke-by-stroke basis. This approach offers a path toward unobtrusive user identification in real-time creative workflows.
Introduction
In collaborative digital workspaces, such as shared Smart Boards or Tablet PCs, multiple users often contribute to a single document simultaneously. Distinguishing "who drew what" typically requires manual intervention, such as switching profiles or pen colors, which interrupts the creative flow. This paper addresses the problem of unobtrusive user identification by leveraging the unique physical habits—specifically pen pressure and tilt—that individuals exhibit while sketching.
Methodology
The researchers utilized a custom drawing application on a Wacom Cintiq tablet to capture data at every point of a stroke, including coordinates, time, pressure ($P$), and tilt ($Tx, Ty$). The research was structured into two phases:
- User Consistency Study: Six users recorded daily to-do lists over three days to evaluate the stability of their sketching style.
- User Identification Study: Three users provided samples of text and geometric objects. A linear classifier was trained on these samples to identify users based on various feature combinations.
User Consistency
A critical requirement for biometric identification is temporal stability. The study found that users maintain highly consistent physical sketching mannerisms across different days.
As shown in Table 1, the mean values for pressure and tilt remained relatively stable for each user across the three sessions, suggesting that these habits are deeply ingrained.
Results and Key Insights
The identification study revealed that physical mannerisms are potent identifiers even for "micro-strokes," such as the dot on an 'i'.
Tilt as a Primary Identifier
The most significant finding was the dominance of pen tilt over pressure. While pressure can fluctuate based on mood or fatigue, the angle at which a user holds the pen acts as a reliable "fingerprint."
Classification Performance
As detailed in Table 2, the system achieved high accuracy across various configurations:
- Three-User Identification: Using all features, the classifier reached 94.0% accuracy.
- Pairwise Identification: Accuracy for specific pairs, such as Users 1 and 3, reached as high as 99.7%.
- Feature Comparison: "All Tilts" alone provided 93.4% accuracy for three users, whereas "All Pressure" alone dropped to 49.4%, confirming tilt as the primary biometric signal.
Discussion and Future Work
While the pilot study demonstrates the potential of stroke-based identification, the small sample size (3–6 participants) is a primary limitation. Future research should focus on scaling the system to larger groups and implementing more advanced machine learning models like Support Vector Machines (SVMs) to handle increased complexity. Ultimately, integrating this technology into real-time multi-user systems could significantly enhance collaborative digital environments by automatically labeling contributions without interrupting the user's creative process.
Summary updated