Regularized MIMO Decoders

In the Multi Input Multi Output (MIMO) antenna system, it is known that the Linear Minimum Mean Squared Error (MMSE) receiver is equivalent to Tikhonov regularization. Given that, we develop a family of generalized receivers based on regularization with different penalty functions that penalize the received symbols outside the convex hull of the modulating constellation. For illustration purposes we consider two types of penalty functions, the deadzone and infinity norm penalty functions. The proposed decoders have low complexity and can be implemented efficiently using convex optimization algorithms. Simulation results show that the proposed receivers outperform the MMSE receiver by as high as 5-dB at low Signal to Noise Ratio (SNR).