PPG signal reconstruction using a combination of discrete wavelet transform and empirical mode decomposition

Photoplethysmographic (PPG) signals, which are measured by pulse oximeter embedded in a form of wristband, are typically used for measuring heart rates. Such wearable sensors may be used for early detection of abnormal conditions for preventive actions in monitoring individual health. However, it is challenging to estimate heart rates using PPG signals with high accuracy due to the irregular motion artifacts, thus making the estimation of heart rate unreliable. In this paper, we proposed the use of Empirical Mode Decomposition (EMD) followed by Discrete Wavelet Transform (DWT) for noise reduction of the PPG signals. We calculated the heart beat rate per minute (BPM) from the reconstructed PPG signals and evaluated the performance of the proposed method in terms of Absolute Maximum Error (AME) and Mean Sum Error (MSE) with the provided ground-truth BPM computed from ECG signals. We have shown an improvement in the MSE values from 67% of the datasets used in this study. We also analyzed the relationship between the performances obtained based the level of movement intensity which are measured using the accelerometer.