Biofeedback from wearable data

For my Master’s in AI and Big Data, I built an Android app for biofeedback aimed at reducing stress using smartwatch data. The project was finished in September 2024.

From signals to a stress prediction

The app integrates a wearable through the Healbe SDK, collecting heart rate and galvanic skin response. Data streams through a server for preprocessing and storage. A machine-learning model predicts stress from resting heart rate, GSR peaks and EEG alpha rhythms.

Training with feedback

Real-time visualisation helps the user follow the measurements. The home screen picks what to lower, heart rate or stress. A resting screen records a one-minute baseline, the training screen runs the session with animated scenes, and the statistics screen shows current and past values.

Implementation

The Android app uses Kotlin and Jetpack Compose, with Koin, Coroutines and Flow, RxJava, Vico charts, Lottie, Coil and Timber. It targets Android 10 and later.

Explore the public Biofeedback repository.