Single-tap Latency Reduction with Single- or Double- tap Prediction

TL;DR

PredicTaps tackles the persistent single-tap latency on touch surfaces by predicting whether a detected tap is a single tap or the first tap of a double tap using a logistic-regression model trained on OS-provided touch events. By applying a PredicTaps activation threshold (PAT) to decide immediate single-tap execution or awaiting a potential second tap, the method achieves drastic latency reductions (approximately 12 ms on touchpads and 17.6 ms on smartphones) while maintaining usability. The authors validate PredicTaps across three data-collection scenarios (daily laptop use, smartphone lab studies, and in-the-wild smartphone use) and two hardware form factors, reporting high per-user and general accuracies when leveraging top-confidence data, along with positive user feedback and manageable false positives. The findings demonstrate PredicTaps as a practical, high-level optimization that reduces latency without hardware changes, with potential for few-shot personalization and broad applicability across devices and applications.

Abstract

Touch surfaces are widely utilized for smartphones, tablet PCs, and laptops (touchpad), and single and double taps are the most basic and common operations on them. The detection of single or double taps causes the single-tap latency problem, which creates a bottleneck in terms of the sensitivity of touch inputs. To reduce the single-tap latency, we propose a novel machine-learning-based tap prediction method called PredicTaps. Our method predicts whether a detected tap is a single tap or the first contact of a double tap without having to wait for the hundreds of milliseconds conventionally required. We present three evaluations and one user evaluation that demonstrate its broad applicability and usability for various tap situations on two form factors (touchpad and smartphone). The results showed PredicTaps reduces the single-tap latency from 150-500 ms to 12 ms on laptops and to 17.6 ms on smartphones without reducing usability.

Figures (11)

• Figure 1: Processing with the conventional tap-detection method: (A) a single-tap event and (B) a double-tap event. Processing with the proposed method: (C) a single tap is predicted, so the system executes a single-tap event immediately after tap detection; and (D) a double tap is predicted, so the system waits for a subsequent tap.
• Figure 2: Processing method without single-tap latency. A single-tap event is executed immediately after every tap.
• Figure 3: Process flow for determining whether a tap is a single tap or the first tap of a double tap using a trained model (in prediction phase). When the system detects a tap, it predicts whether the tap is a single tap or the first tap of a double- tap. If the system judges the tap to be a single tap with a confidence rate higher than PAT), a single-tap event is executed immediately without the single-tap latency.
• Figure 4: Accuracy and ROC curves for different percentages of data used for prediction in the uncontrolled experiment for a laptop touchpad. The red line in (A) is the accuracy of the general model trained with all participants' data, and the other lines are of the individual models. (B) is the ROC curve of the general model for different percentages of data used according to their calculated confidence scores ( PAT).
• Figure 5: Screenshots of our experiment application. A) Annotation Task part. The options are goat, dolphin, hippopotamus, and donkey. B) Pointing Task part. C) Instruction part. The instruction says, "In this session, you need to tap five times. Double-tap the rectangles on the screen. Press the 'next' button if you are ready."