Adaptive system identification using interior point optimization

We present a new algorithm for the adaptive estimation of the tap weights of an unknown linear transversal filter. This algorithm takes advantage of the fast convergence properties of some recently developed interior-point optimization techniques. In particular, we use ideas from interior-point column generation methods, whose iterative nature lends itself well to applications that require adaptive solutions. Numerical simulations demonstrate that the new algorithm compares well against RLS, in terms of convergence speed, especially when conditions are adverse (i.e., SNR is low, input signal is correlated, systems are time-varying).