Bit-Constrained Least Squares for Improved Detection

We consider the problem of detecting multiple communication signals in the presence of noise, where the vector of received signals is a linear combination of the symbols of the individual signals. This problem arises for example in MIMO systems employing multiple antennas at both the receiver and transmitter. The maximum likelihood detector offers an optimal but computationally expensive solution for this problem. Least-squares methods require only a modest amount of computation but may have relatively poor performance. We propose a bit-constrained least-squares algorithm whose computational complexity is comparable to that of least-squares methods, yet its performance is significantly improved.