First Order Adaptive IIR Filter for CQI Prediction in HSDPA

In this work, we consider channel quality indicator (CQI) prediction for High Speed Downlink Packet Access (HSDPA). In HSDPA systems, there is a delay of about 3 subframe associated with the application of CQI feedback. Traditional FIR filter based predictors such as Least-Mean-Squares (LMS) and Recursive-Least-Squares (RLS) typically come with high computational complexity. We consider a first order adaptive IIR alternative with substantially lower complexity while attaining the same level of accuracy. By minimizing the mean squared error (MSE), we derive an exact gradient descent algorithm as well as two pseudolinear regression algorithms. The convergence rates and the convergence conditions are then established. Simulation results show that the proposed adaptive IIR filters combat both feedback delay as well as estimation error and provide up to 25% HSDPA throughput improvement.