A low cost coupled LMS adaptive algorithm with affine projection performance for autorregressive inputs: Mean weight analysis

This work presents a low cost fast convergence adaptive algorithm for autoregressive (AR) input signals. It basically consists on an NLMS adaptive algorithm coupled with a linear adaptive predictor. Mean weight theoretical analysis shows that this adaptive structure, under AR conditions, provides a performance quite similar to the well known Affine Projection algorithm. Theoretical results are supported by Monte Carlo simulations showing a very good matching for considerably correlated signals. This adaptive algorithm could be a good choice in control and filtering applications that require high adaptation speed and low computational cost.