Signal Processing For Local Pulse Wave Velocity Estimation Based On Fourier Decomposition

Pulse Wave Velocity (PWV) is the optimal index for quantitative evaluation of arteriosclerosis. Owing to the transmission of carotid pulse wave (PW) signals are affected by reflected wave, and the quantization errors occur when moving cross-correlation is used to estimate the pulsation displacement, the accuracy of Time Transit (TT) method in estimating local PWV is decline. For improving the accuracy of PWV detection, a smoothing estimation method based on Fourier-based Filter (FT-Filter) is proposed in this paper. Fourier transform is used to transfer the PW from time domain to frequency domain, and the filter parameters are determined by the spectrum characteristic to filter the high frequency components. Then the time domain denoised PW is obtained by inverse transform. The proposed method is used to process the simulated and clinical PW transmission sequences. The PWV is estimated by TT method, and compared with the estimated results of Savitzky-Golay and Butterworth Filter, the normalized root mean square errors of 30 simulated cases are $0.28\pm 0.05,\ 0.54 \pm 0.07$ and $0.39\pm 0.06$, respectively. The clinical results of 20 subjects' PWV are 4.51 m/s, 5.86 m/s and 5.19 m/s, respectively. It shows that the proposed FT-Filter improves the accuracy of local PWV estimation based on TT method. There will be helpful to improve the accuracy of clinical diagnosis of local arteriosclerosis.