Signal Processing Methods for Heart Rate Detection Using the Seismocardiogram

Abstract

Cardiac diseases are one of the major causes of death. Heart monitoring/diagnostic techniques have been developed over decades to address this concern. Monitoring a vital sign such as heart rate is a powerful technique for heart abnormalities detection (e.g., arrhythmia). The novelty of this work is that offers new heart rate detection methods which are both robust and adaptive compared to existing heart rate detec- tion methods. Utilized data sets in this research have been provided from two sources of PhysioNet and a research group. In this work, utilized methods for heart rate detection include Signal Energy Thresholding (SET), Empirical Mode Decomposition (EMD) and Empirical Wavelet Transform (EWT). To the best of the author’s knowledge, this work is the first to use EMD and EWT for heart rate detection from Seismocardiogram (SCG) signal. Obtained result from applying SET to ECG signal is selected as our ground truth. Then, all three methods are used for heart rate detection from the SCG signal. The average error of SET method, EWT and EMD respectively 13.9 ms, 13.8 ms and 16 ms. Based on the obtained results, EMD and EWT are promising techniques for heart rate detection and interpretation from the SCG signal. Another contribution of this work is arrhythmia detection using EWT. EWT provides us with the instantaneous frequency changes of the corresponding modes to ECG signal. Based on the estimated power spectral density of each mode, power spectral density of arrhythmia affected ECG is higher (≥ 50dB) compared to the power spectral density of a normal ECG (≤ 20dB). This provides the potential for arrhythmia detection using EWT.