Eye State Prediction on Android Devices using Machine Learning for Natural Environment Electroencephalogram...

Electroencephalogram (EEG) signals capture brain activity and are valuable for cognitive and medical applications. While deep learning models achieve high EEG classification accuracy, they require large datasets and significant resources. To overcome this, we developed a lightweight machine learning pipeline for EEG classification on Android devices using TensorFlow Lite, optimized for limited data. As a case study, we classified eye states (eyes open vs. eyes closed) using EEG signals collected from ten participants via the CameraEEG app. After artifact removal with Embedded-ASR and power spectral feature extraction, a single-channel Support Vector Machine (SVM) model was trained and deployed. The model achieved 90% accuracy and showed robustness across multiple evaluation metrics. Tested on smartphones like Google Pixel 7 Pro and Samsung S22, the Android app performed EEG classification. This work advances physiological measurement in natural settings and can extend to cognitive workload monitoring, seizure detection, and mental health assessment.