Abstract:
In this thesis, spectral estimation methods are applied to the variation in the time interval between heart beats called heart rate variability (HRV) and further to estimate the degree of accuracy of diagnosis according to signals. Frequency domain models are analyzed using four methods of spectral estimation. Fast Fourier Transform (FFT), a specialized technique for frequency domain analysis is first applied on the data. The methods of spectral estimations were then applied and the variance of their spectrum was considered. The sample periodogram has a variance of 465.99, singular spectrum analysis (SSA) with variance 1.0020, maximum entropy method (MEM) with variance 0.4951, as well as multitaper method (MTM) with the smallest variance of 0.00075352. Evaluation is performed by observing the behavioral pattern of the signals produced by each spectrum plot and the best spectrum is determined by calculating the bias and variance of each spectrum. The results has established that spectral analysis using multitaper method which has the spectrum with the minimum variance is an efficient and useful approach in heart rate signal analysis, it is also shown that the multitaper spectra method with the minimum variance is the best method of estimating heart rate variability at rest.
