A noise-resilient affine projection algorithm and its convergence analysis

• A new constrained criterion using past estimates is proposed.
• The proposed NR-APA has a lower steady-state error as the number of past estimates increases.
• This approach is more effective at very low SNR levels.
• Theoretical performance analysis describes well the simulation results.

Recently a new normalized least mean square algorithm has been proposed by minimizing the summation of the squared Euclidean norms of the changes between the weight vectors to be updated and the past weight vector. The resultant algorithm exhibits noise resilience in that they prevent the adaptive filter from fluctuating around an optimal solution, but its convergence behavior has not been studied in detail. Thus, we first apply the constrained criterion to an affine projection algorithm (APA) for identifying a highly noisy system by reusing weight vectors. Since the performance of the APA declines under low signal-to-noise ratio (SNR) conditions, this approach is more effective for decreasing the steady-state mean-square deviation (MSD). Then, we analyze the convergence behavior of the proposed APA theoretically using energy conservation arguments. The experimental results show that the proposed theoretical results agree well with the simulation results.