NanoSleep AI Classifies Sleep Stages Efficiently

This research introduces NanoSleep, a highly compact and efficient hybrid temporal convolutional network designed for automatic sleep stage classification using single-channel electroencephalography (EEG) data. This capability is crucial for the development of wearable and home-based sleep monitoring devices, which often operate under significant resource constraints. NanoSleep's architecture integrates several key components: a learnable Sinc-convolutional front end, a dual-branch feature extractor that combines multi-scale temporal and spectral representations, a gated dilated temporal convolutional backbone with channel recalibration, and a conditional random field for sequence-level decoding. To address class imbalance, a weighted calibrated focal loss is also employed. Evaluated on the Sleep-EDF and Sleep-EDF-Expanded datasets, NanoSleep consistently outperformed six established baseline methods. An ablation study confirmed the individual contribution of each component to its overall effectiveness. These results highlight NanoSleep's ability to balance high accuracy with computational efficiency, making it an ideal solution for deployment on wearable devices and in resource-limited clinical settings.