Energy-Efficient FPGA Based Sleep Apnea Detection Using EEG Signals.

Sleep apnea is a prevalent sleep disorder characterized by frequent interruptions in breathing during sleep, leading to decreased levels of blood oxygen. This research introduces an energy-efficient digital hardware system built on an Artix 7 FPGA, explicitly designed for real-time sleep apnea detection. Our approach involves the classification of subject-specific sleep apnea and non-apnea events. We utilize inter-band energy ratio features extracted from multi-band Electroencephalogram (EEG) signals and employ a Linear Support Vector Machine (LSVM) classifier for this task. The features extracted—namely energy, kurtosis, and mobility—from five sub-bands demonstrate improved accuracy, sensitivity, and specificity compared to existing studies. The proposed model is evaluated using EEG signals from the openly accessible St. Vincent’s sleep apnea UCDDB database. Our system achieves remarkable performance metrics, attaining the highest accuracy of 94.81%, a sensitivity of 93.10%, and a specificity of 96.43%. It accomplishes all this while maintaining minimal dynamic power consumption (19mW) and using minimal FPGA resources. This hardware system can be integrated into a System-on-a-Chip (SoC) platform, serving as a crucial component of a smart, wearable, automated sleep apnea detection device for real-time critical health diagnosis and screening.