Nonlinear adaptive digital filter with sequential regression (SER) algorithm

A class of adaptive nonlinear filtering algorithms are discussed in this thesis. The algorithms derived in this thesis use the sequential regression (SER) method to compute the matrix inverse in a recursive way. Assuming Gaussian signals, the algorithms update the second order Volterra filter coefficients. Among those algorithms, long-term equally weighted SER algorithm is suitable for adaptive filtering applications related to stationary signals and the short-term exponentially weighted SER algorithm is suitable for those in nonstationary environments. Computer simulation results pertaining to SER algorithm are presented. For the purpose of comparison, corresponding result using least mean square (LMS) algorithm is also included.