Energy-efficient ECG compression in wearable body sensor network by leveraging empirical mode decomposition

Wearable body sensor network (BSN) is widely used in daily monitoring, well-being management, and rehabilitation. Energy efficiency imposes a stringent constraint in wearable BSN, in which wireless transmission is significantly power-demanding. Compressed sensing (CS) provides a good solution to reduce power consumption for data transmission due to the sparsity of signals which can use limited transmitted data to reconstruct original signals. In this study, we develop a new method for non-sparse ECG signal compression by leveraging empirical mode decomposition (EMD) and online dictionary for wearable devices. Comparing to the state-of-the-art of ECG compression which can achieve the compression ratio (CR) of around 25 with the root mean square error (RMSE) around 5%, our method can achieve the CR up to 60 with the same level of RMSE for wearable ECG. In addition, our method also has low computational complexity, which can achieve lower compression energy. The validation experiments are conducted on both clinical data and wearable ECG detected by our BSN in noisy environment. The proposed method shows high feasibility for real CS on board to achieve ultra-low power consumption