TouchML

touch targeting models for data analysis, Android, and web

TouchML:
Model targeting behaviour

Targeting experiments help HCI researchers and practitioners to quantify interactions - for example by analysing speed/accuracy according to Fitts' Law. In contrast, touch offset models capture spatial patterns of touch inaccuracy across the screen. This enables new analyses and applications.

Improve touch accuracy

Touch offset models improve touch accuracy: Modelling targeting behaviour, devices can predict the user's true intended touch locations based on imprecise sensed locations [1, 3, 4].

153.600 touches

TouchML comes with a large database to start training and applying models right away. In a controlled lab study, we collected 4 target types × 400 target locations × 2 hand postures × 24 participants × 2 sessions = 153.600 touches.

Explore further

Touch offset models have also been explored to recognise users and hand postures [1], to reveal characteristic screen regions for describing individual touch targeting behaviour [4], and to reduce typing errors [5].

Examples
Documentation

What will you build?

Further Reading

Check the paper's related work. Here are some direct links as well:
1. Buschek et al. 2013 | 2. Henze et al. 2011 | 3. Weir et al. 2012 | 4. Weir et al. 2013 | 5. Weir et al. 2014