HandiSense: Left/Right Operating Hand Recognition for Multi-finger Touchscreen Interactions

Touch-based interfaces dominate today's mobile and embedded devices, creating a need to accurately distinguish the operating hand (left vs. right) behind touch inputs for personalized, context-aware interaction. We present HandiSense, a purely software solution that determines the operating hand using only the raw multi-touch data exposed by standard capacitive-touchscreen APIs—no cameras, wearables, or low-level sensor access required. HandiSense captures distinctive geometric and kinematic signatures by constructing spatial and temporal triangles from concurrent touch points and their trajectories. A modular classification framework is first trained on two-finger gestures and then extended to handle up to five fingers through masked feature fusion. Evaluated on over 70,000 multi-finger samples from diverse users, devices, and scenarios, HandiSense achieves over 96% accuracy and maintains stable performance across varying screen sizes and interaction styles. By offering a calibration-free, cross-device, and wholly software-based path to hand-aware interaction, HandiSense broadens the design space for personalized and context-adaptive touch interfaces.