Abstract
Wearable sensors and IoT technologies have enabled real-time monitoring of vital signals and created new perception of medical patient-care strategies. Atrial fibrillation (AF) as the most common cardiac arrhythmia that is associated with increased risks of strokes can be detected in ECG signals using wireless body sensors. In depth, these devices can be deployed to capture and transmit raw ECG signals to a remote server for AF detection and further ECG classification. However, the transmission of raw ECG samples over energy-consuming wireless links limits the sensor's battery lifetime and questions the application's practicality. The design of an on-sensor AF energy efficient detection method is still an open research issue challenged by the sensor processing capability and its lifetime. This paper proposes a new ECG processing scheme for AF detection intended to be implemented in WBSN. It presents a low-complexity time-domain features' extraction method that combines the RR interval and p-wave absence in the ECG segment for higher AF detection accuracy. The proposed method achieved 100% sensitivity and 96% specifity while saving 92% of energy compared to full ECG data transmission. Thus demonstrating the technical feasibility of energy-efficiency for on-sensor AF detection.