On joint detection and channel estimation over rank-deficient MIMO links with sphere decoding

In this paper, we present a method which jointly performs robust channel estimation and signal detection over rank-deficient MIMO links. This technique achieves an excellent estimation of the channel matrix by using all detected symbols within a data frame. Symbol detection is achieved by means of a sphere decoding algorithm. The proposed scheme is equivalent to having the training preamble occupy the entire data frame. The Mean Square Error (MSE) of the channel estimation for this method is close to 10^-3 for 16-QAM modulation over a wide range of preamble lengths. The approach is applicable in both full rank (N = M) and rank-deficient (N <; M) channels with practically the same performance as that of a clairvoyant detector with perfect CSI knowledge. This characteristic is crucial when co-channel signals are sent over an underdetermined channel, where there are more antennas at the transmitter than there are at the receiver, i.e. virtual MIMO processing.