Fig. 5
From: Systolic blood pressure estimation method using electrocardiogram RRI data
Flowchart for determining the frequency ranges of LSR and USR. By using the \(\:{\theta\:}_{i-0}\) spectrum (1), the IPCM approximation curve (2) was expressed with a quintic function in the frequency range between 0.006 and 0.06 Hz. From the quintic function, IPs (IP1 to IP3) and intersections of tangent lines of IPs (IT1, IT2) were described (2). The vicinity of IP1, IT1 and IT2 having small residuals were active parts where the robustness of phase was high, while the vicinity of IP2 and IP3 having large residuals were defined as regulation sites where the plasticity of phase was high. The interval between IT2 having a small residual and IP2 having a large residual is free site8 linking the two sites. Accordingly, by using the residuals between the approximation curve of the quintic function and spectra, an IPCM residual function diagram of a quintic function with residuals as the vertical axis and frequencies as the horizontal axis was created (3). For the residual function composed of the frequency range between 0.006 and 0.06 Hz, there occurred a stationary point in the vicinity of IP2 and the local maximum value in the vicinity of IP3. As a result, IP2 and IP3 become the maximum trajectory frequencies. Successively, using the least squares method, residual regression curves (residual approximation curves) were created to classify the residuals into two clusters. A regression curve created with the cluster having larger residuals than the regression curve was defined as the high group of the residual probability distribution curve, while a regression curve created with the cluster having smaller residuals was defined as the low group (3). Incidentally, the residual regression curves were created by making regression lines with the linear axis and expressing the frequencies on the horizontal axis with the logarithmic axis. The vicinity of the stationary point of the IPCM residual function became the maximum trajectory frequency of LTM, and the vicinity of the local maximum value became the maximum trajectory frequency of STM (4)
