PreTap: Implicit Reachability Analysis and Tap Prediction for One-handed Mobile Interaction

One-handed interaction on large-screen smartphones often leads to ergonomic discomfort and disruption, while existing solutions typically rely on explicit gestures, additional hardware, or static ergonomic models, limiting their deployability and adaptability. We present PreTap, an anticipatory system that implicitly infers reachability and predicts imminent tap regions from zero-permission inertial signals, enabling proactive interface adaptation during the pre-tap phase. Realizing this capability introduces challenges, including the extremely short prediction horizon, substantial inter-user variability in reachability zones, and IMU signal patterns. To address these challenges, PreTap combines an on-device (i.e., 180K-parameter) CNN-Attention model with an unsupervised reachability analysis algorithm that captures user-specific ergonomic constraints and supports user-specific adaptation goals. An empirical study with 50 participants on commodity smartphones shows that PreTap can accurately predict the tap intent and proactively adapt the UI approximately 220 ms before contact, while maintaining low runtime overhead suitable for real-time deployment. These results demonstrate the potential of implicit motion sensing to enable intelligent, reachability-aware mobile interfaces that better align with users' motor intent.