No Needles Attached? Inferring Energy Metabolism Zones and Lactate Accumulation from Touchscreen Input

Dominik Schön, Uwe Gropengießer, Max Mühlhäuser, Florian Müller

2026

Proceedings of the Augmented Humans International Conference 2026

TL;DR

What we did: We designed a method to non-invasively infer Energy Metabolism Zones and Lactate Accumulation using smartphone touchscreen input combined with built-in sensors.

What we found: We found that our machine learning models can accurately classify physiological states related to lactate levels based on minimal interaction, with better performance observed when using participant-specific data.

Takeaway: Our work demonstrates the potential for smartphones to act as low-cost, non-invasive tools for athletes to monitor training intensity, suggesting significant implications for personal fitness and health management.

Abstract

Recreational athletes increasingly adopt quantified-self practices to track and advance their training, recovery, and fitness. Blood lactate is a key biomarker in this context, but testing remains invasive and costly, limiting its use to professional sports and clinics. For everyday exercisers, even a coarse distinction, such as whether they are training below or above key thresholds, already provides actionable insight. We investigate whether commodity smartphones can classify these thresholds non-invasively using swipe input and built-in sensors. In a data-collection study, participants completed touchscreen tracing tasks at varying physiological states during their workout, while collecting blood samples. We analyzed touch, pressure, motion, and task features to understand their role in classifying the Energy Metabolism and Lactate Accumulation Zones and trained a Support Vector Machine and a Recurrent Neural Network. Our results demonstrate the feasibility of estimating these zones from short smartphone interactions, suggesting a path toward accessible, non-invasive on-device training guidance.