Low complexity detection for massive MIMO under multipath fading with limited storage resources

For massive MIMO system, few research has been done for multipath frequency selective fading model among all literatures. Hence, in this paper we propose a 2-stage multibranch linear MOE (minimum output energy) receiver which collects symbols from different paths to deal with frequency selective fading environment. Due to the big number of antennas, we are using sample covariance matrix other than estimating all channels. To avoid collecting large number of samples (which requires additional storage resources and may cause significant system delay), we are focusing on detection methods which use less samples. With the aid of a new method of singular matrix preconditioning and efficient series expansion, we are able to show that this implementation not only achieves lower computational complexity but also outperforms the conventional reduced-rank receiver and the single branch MOE receiver as well.