Heart Rate Variability Analysis With Frequncy Domain Methods & Artificial Nerual Network

Heart rate variability (HRV) is the fluctuation of beat-to-beat intervals, and research for it is important in clinic. The new developed new signal processing methods and Artificial Neural Network (ANN) theory provide new research methods HRV. Wavelet transforms , autoregressive (AR) model power spectrum density (PSD) estimates and ANN are used to analyze HRV signals in this paper, in order to get valuable information and provide new diagnosing means. Several researches were made, which briefly described below.1. AR model PSD estimates in 30mins HRV signals were studied in this paper. Frequency domain indexes of Congestive Heart Failure (CHF)patients with normal are compared. Results show severe CHF frequency indexes of HRV total power, powers of different frequency bands,peak values of low and high frequency are all decreased. 5 minutes short term HRV signals are analyzed and there are significant differences between severe CHF and normal.2. Wavelet Analysis in 30mins & 5mins HRV were studied in this paper. Wavelet Multi-resolution Analysis method was used to decompose HRV into several frequency band, and then the frequency domain indexes were obtained. Compare AR model PSD estimates with wavelet analysis in indexes obtained from HRV. Frequency domain indexes of CHF patients with normal are compared. Results show the indexes obtained by wavelet can clearly describe the difference between severe CHF and normal.3. Construct ANN for CHF diagnosis. Great performance was presented when use indexes from AR model PSD estimates and wavelet analysis as the inputs of ANN. It means that both two indexes can definitely describe the difference between severe CHF and normal.4. Design HRV analyzing system. Built a analyzing system based on MATLAB environment. The system has friendly dialog interface between persons and machine, and it is easy to use. It mainly task is to support ANN auto-diagnosis based on HRV.