This paper addresses trained adaptation of broadband multiple-input multiple-output (MIMO) equalisers operating in severely time-dispersive environments which result in slow convergence and poor performance for least mean squares (LMS) type algorithms. We present a subband approach, which enhances a normalised LMS (NLMS) convergence by prewhitening of the input signals and reduces its computational complexity. Equalisation based on the recursive least squares (RLS) algorithm in the time domain is prohibitive due to its high complexity, which can be drastically reduced by the subband approach to an extend such that the overall RLS cost for convergence may be much lower than the subband NLMS. The performance of the various systems is compared in simulations.
展开▼
