Fast Array Multichannel 2D-RLS Based OFDM Channel Estimator

In this paper a Fast Array Multichannel Two-Dimensional Recursive Least Square (FAM 2D-RLS) adaptive filter is proposed for estimating an OFDM channel in frequency domain. This filter makes use of the shift structure of the input data vector. Thus the computational cost of the classical RLS filter which is O(M) is reduced to O(M) for each iteration where M is the order of the filter. In order to ensure numerical stability in finite precision, we make use of array based methods for implementing FAM 2D-RLS. The adaptive filters illustrated in the standard literature consist of a weight vector and a scalar desired data. But in our scenario of OFDM channel estimation the weight is a matrix while the desired data is a vector. Hence the algorithm for matrix form of FAM-2D RLS and its steady state equations are derived. Numerical stability, steady state and convergence performance are verified using MATLAB simulations.