2. Random signal models

Signal models involving random quantities are often used. There are two main reasons for this:

• the need to model relatively large classes of signals, grouped by certain generic properties (e.g. audiophonic signals, the speech signal, images, etc.);

• the need to model various types of "noise" (e.g. measurement noise), which are generally difficult to control.

The mathematical framework adapted to this situation is that of random processes. We won't go into probabilistic details here (a more detailed discussion can be found in