Notes on Digital Signal Processing: Practical Recipes for Design, Analysis and Implementation

Note 68. Autoregressive Signal Models

Many real-world signals can be modeled as autoregressive (AR) processes. The properties of AR processes have led to the development of numerous techniques for analyzing and characterizing of such signals. This note serves as an introduction to these techniques, which are explored in subsequent notes.

An autoregressive process of order p (often denoted as an “AR(p) process”) can be generated using a p-stage all-pole filter driven by a white noise source, as shown in Figure 68.1. The configuration shown in the figure implements the difference equation (68.1) given in Math Box 68.1, and is sometimes referred to as an autoregressive signal model.

Figure 68.1. All-pole filter configured for generating an AR( ...