The Delineation of Aortic Valve Opening Point and Estimation of Heart Rate Variability from Seismocardiogram signal using Linear Prediction Coding Technique

Our study presents linear prediction coding (LPC) technique-based AO peak detection framework for seismocardiogram (SCG) signal and heart rate (HR) estimation. In the proposed method, residual signal is estimated by inverse filtering of SCG signal. This residual signal emphasizes the systole and diastole profiles of the SCG signal. The AO peaks of the systole profiles are detected using Hilbert transform and first-order gaussian difference (FOGD) filter. The proposed method is tested on the CEBS database from the PhysioNet archive. The best overall performance for detecting AO peaks is found to be precision of 96.23%, sensitivity of 97.78%, and accuracy of 94.17%. For robustness of the proposed method, Monte Carlo simulation is performed for the noise stress test by adding additive white gaussian noise (AWGN) to the SCG signal. The proposed method can achieve good performance under this simulation. Further, the heart rate variability (HRV) is estimated using our proposed method and compared with the HRV evaluated from the reference ECG signal. The proposed method can achieve performance close to the reference value.