Classifier based low-complexity MIMO detection for spatial multiplexing systems

In this paper, we propose a low complexity detection scheme for MIMO systems incorporating spatial multiplexing. Optimal detection schemes such as maximum-likelihood (ML) detection of MIMO signals demands computational resources that are beyond the capabilities of most practical systems. Alternative reduced complexity MIMO detection techniques have been proposed, but the complexity of algorithmic schemes are in general much higher than that of the equalizer-based techniques, e.g. zero-forcing (ZF) or MMSE. On the other hand, equalizer-based techniques perform relatively poor in terms of error rate. In this paper, we propose a hybrid of an equalizer-based technique and an algorithmic search stage. Based on an error matric and its probability density functions for different classes of error, a particular search region is selected for the algorithmic stage. As the probability of occurrence of error classes with larger search regions is small, overall complexity of the proposed technique remains low while providing a significant improvement in the error performance.