Effect of non-uniform sampling on biomedical signals

Heart Rate Variability (HRV) is one of the reliable quantitative markers of the Automatic Nervous System activity. It can be analyzed in time and frequency domain. This report will focus on frequency domain analysis. Usually HRV is divided into four bands in frequency domain. Among them Low Frequency power; LF (0.04 – 0.15 Hz), High Frequency power; HF (0.15 – 0.4 Hz) and also LF/HF – ratio are most often used frequency domain measures. In general, HRV consists of an unevenly spaced time series. Lomb-Scargle Periodogram (LSP) is proven to be the best method for unevenly sampled time series. Conventional methods such as Welch and Bartlett’s methods of power spectral estimation needs evenly sampled time series which leads to re sampling of HRV time series to evenly spaced time series. This report deals with Lomb-Scargle Periodogram estimation, different interpolation techniques to resample HRV time series prior to compute Bartlett and Welch’s methods using FFT and also noise filtering, peak detection algorithm, ectopic beats elimination from RR tachogram. HRV analysis comprises of comparison of HRV for different aged persons and also effects of aging, ectopic beats on power spectral components are investigated in this report. ECG signals for the analysis are collected from MIT-BIH database. This database contains mostly abnormal or unhealthy electrocardiogram signals that can be used for research purpose. At last, the HRV analysis for the MIT-BIH database is compared with using Integral Pulse Frequency Modulation (IPFM) model. Developed matlab algorithms for the project are discussed in Appendix A.Master of Science (Signal Processing