Energy-Efficient Real-Time 4-Stage Sleep Classification at 10-Second Resolution: A Comprehensive Study

--Sleep stage classification plays a crucial role in health monitoring, particularly for diagnosing and managing sleep disorders such as sleep apnea and insomnia. However, conventional clinical approaches like polysomnography are often costly, inconvenient, and impractical for long-term, home-based monitoring. In this study, we present an energy-efficient classification approach for detecting four sleep stages--wake, rapid eye movement (REM), light sleep, and deep sleep--using a single-lead electrocardiogram (ECG) signal. We evaluate and compare the performance of various machine-learning and deep-learning models. T o support this, we introduce two novel windowing strategies: (1) a 5-minute window with 30-second steps for machine-learning models utilizing handcrafted features, and (2) a 30-second window with 10-second steps for deep-learning models, enabling near-real-time predictions with 10-second temporal resolution. Although lightweight, our deep-learning models--such as MobileNet-v1--achieve high classification performance (up to 92% accuracy and 91% F1-score), their energy demands remain high, making them sub-optimal for wearable applications. T o address this, we design a SleepLiteCNN optimized specifically for ECG-based sleep staging. T o further enhance efficiency, we apply 8-bit quantization, which leaves classification performance unchanged, reducing the energy usage of our SleepLiteCNN to just 5.48 µJ per inference at a 45 nm technology node, with 90% accuracy and 90% F1-score. We further demonstrate that deploying this SleepLiteCNN on a field-programmable gate array (FPGA) significantly reduces resource usage through quantization. Overall, this approach provides a practical and efficient solution for continuous ECG-based sleep monitoring in compact, resource-constrained wearable devices. LEEP Sleep stage classification is crucial in health monitoring, especially for diagnosing and managing sleep disorders such as sleep apnea and insomnia [2]. Sleep stages are categorized into wake, REM, and Non-Rapid Eye Movement (NREM) stages. Z. Mohammadi is a Ph.D. Candidate in the School of Electrical and Computer Engineering, University of Tehran, Tehran, Iran (e-mail: zahramo-hammmadi@ut.ac.ir). Fazel is a Graduate of the School of Electrical and Computer Engineering, University of Tehran, Tehran, Iran (e-mail: parnian.fazel@ut.ac.ir).