Adaptive linear filtering using interior point optimization techniques

We propose a novel approach for the linear adaptive filtering problem using techniques from interior point optimization. The main idea is to formulate a feasibility problem at each iteration and obtain as an estimate a filter near the center of the feasible region. It is shown, under some mild conditions, that this algorithm generates a sequence of filters converging to the optimum linear filter at the rate O(1/n), where n is the number of data samples. Furthermore, we show that the algorithm can be made recursive with a per-sample complexity of O(M^2.3), where M is the filter length. The potential of the algorithm for practical applications is demonstrated via numerical simulations where the new algorithm is shown to have superior transient behavior and improved robustness to the source signal statistics when compared to the recursive least squares (RLS) method