Fast and Accurate Power Spectral Analysis of Heart Rate Variability using Fast Gaussian Gridding

Abstract

In this paper, we propose an algorithm for fast and accurate Power Spectral Analysis of Heart Rate Variability based on the Lomb Periodogram. The previously introduced Fast-Lomb periodogram, may have reduced the computational complexity of PSA, however it still requires a large oversampling factor, which increases the complexity of the needed FFTs. In our approach, by utilising the Fast Gaussian Gridding method we produce accurate evenly spaced grids for the required FFTs by restricting the oversampling factor only to 2. By doing so, the required FFT size is reduced by up to 4 times without compromising the output accuracy. Our results indicate that the proposed spectral analysis system can achieve upto 76.55% savings in the number of operations or up-to 75.8% in terms of the total execution time.