Adaptive power spectrum estimation algorithm for heart rate variability analysis

The authors present an online, adaptive method of heart rate variability (HRV) spectrum estimation which extends readily to the real-time analysis of time-varying cardiovascular systems. An autoregressive (AR) model is assumed. The AR parameter vector is estimated using a recursive least squares (RLS) method which incorporates an online tuned robustness factor to dampen the influence of ectopic heart beats. The roots of the associated AR polynomial are found using a recently described adaptive polynomial factorization method (D. Storer and A. Nehorai, 1991). Spectral frequency peaks and powers are estimated from the residues of the associated complex function. This method of adaptive HRV spectral estimation has applications in cardiovascular investigations and physiologic monitoring