Estudo e aplicações de filtragem digital adaptativa utilizando o algoritmo LMS

Abstract

This work approaches the study of adaptive digital filters. The main feature of an adaptive filter is the ability to automatically and continuously adjust a given signal to obtain a desired response. This allows applying it in cases when a fixed coefficients filter is not suitable. The adaptation of the filter coefficients are computed using some error estimat ive, which are derived from a reference and input signal. To control this process, adaptive algorithms are used, such as LMS, which minimizes an objective function based on the mean square error. Based on the study of adaptive filters and adaptation algorithms, applications are developed that may illustrate the design methodology for each of the four basic classes of applications: system identification, inverse modeling (or adaptive equalization), signal prediction and signal enhancement (or interference cancellation). These classes of applications are differentiated from each other by the way the reference signal is characterized and extracted. To validate the results, simulations are performed in Matlab ®, described according to a mathematical model developed for each application.