This document discusses the Wiener filter, which is used to estimate an original signal x(n) from a distorted signal y(n). It presents the basic formulation of the Wiener filter, which minimizes the mean square error between the estimated signal x(n) and the original signal by calculating the filter coefficients h(i). It also discusses modeling the distorted signal y(n) using different models, such as a linear model where y(n) is a linear transformation of x(n) with added noise, and convolution models. This allows calculating the filter coefficients without directly observing the original signal x(n).