Motion Artifact Removal in R-R Interval Estimation of ECG Signal for Wearable Biomedical Sensors Using Empirical Mode Decomposition and High-Pass Filter

Clothing-type wearable electrocardiogram (ECG) sensors are effective for continuous long-term monitoring because they impose minimal burden on the body. However, these sensors are susceptible to noise (motion artifacts) caused by body movements. As a result, estimation accuracy may become unstable when the electrodes are not firmly attached to the skin. An adaptive filter using an accelerometer has been proposed as one method to remove motion artifacts. Nevertheless, biomedical sensors and accelerometers are not always integrated, and the use of a separately installed accelerometer complicates synchronization and increases overall system cost. A visual inspection of the frequency spectra of motion artifacts and ECG signals revealed that the ECG contains a fundamental frequency of 1.3 Hz along with its odd-numbered harmonics, whereas the frequency components of motion artifacts are concentrated primarily below 4 Hz. Based on this observation, we developed a new high-pass filter (HPF) design method. Although empirical mode decomposition (EMD) is effective for noise removal, EMD alone is insufficient to eliminate significant motion artifacts in wearable devices. Therefore, we propose a new configuration that combines the designed HPF with EMD.