Tutorial on univariate autoregressive spectral analysis.

Area of Science:

• Signal Processing
• Biomedical Engineering
• Time Series Analysis

Background:

• Spectral analysis reveals frequency content and variation sources in time series.
• Autoregressive (AR) modeling offers an alternative to discrete Fourier transform for spectral estimation.
• AR modeling is particularly valuable for high-resolution spectral estimation of short time series.

Purpose of the Study:

• To explain the theoretical basis of autoregressive (AR) modeling in spectral analysis.
• To demonstrate the application of AR modeling in biomedical engineering, specifically for heart rate variability and electroencephalogram tracings.
• To illustrate AR spectral analysis using heart rate variability data.

Main Methods:

• AR modeling involves regressing each time series value on its preceding values (model order).
• The AR model functions as a filter, separating a time series into predictable and prediction error components (AR analysis filter).
• The AR model can be inverted, treating the prediction error as input and the time series as output (AR synthesis filter) to analyze frequency components.

Main Results:

• The AR analysis filter decomposes time series into predictable components and prediction errors.
• The AR synthesis filter's properties allow for the determination of amplitude and frequency of time series components.
• The study illustrates the practical application of AR spectral analysis with heart rate variability data.

Conclusions:

• Autoregressive modeling is a powerful technique for high-resolution spectral analysis, especially for short time series.
• Its application in biomedical engineering, particularly for analyzing heart rate variability and EEG, is significant.
• Understanding the theoretical basis and filter properties of AR modeling enhances its utility in signal processing.