On utilizing the strengths of ar and MVDR methods to circumvent their weaknesses in application to HRV analysis

We discuss a procedure that exploits the strengths of autoregressive (AR) and minimum variance distortionless response (MVDR) methods to circumvent their respective weaknesses, in the context of estimating the powerspectral measures from cardiovascular signals. While the AR power specfrum density (PSD) has high resolution, it suffers from a greater degree of variability at low orders and potential spurious peaks at high orders. The MVDR method (of lower resolution) provides the power-spectral measures at lower variabiliq over low filter orders. The weaknesses of the two methods can be circumvented by estimating the powers by Low-order MVDR filters at the frequencies estimated by AR PSD. Results of experiments with simulated and real data are supportive.