Low-complexity processing of randomly correlated signals
Christian B. Schlegel, Zhenning Shi, Zachary Bagley · 2003
edu Abstract - Linear pre-processing of systems with large numbers of concurrent channels such as asyn- chronous CDMA systems or MIMO systems is used to transform the multi-access channel into parallel single-user channels. The linear filters such as decor- relator and MMSE filters typically have a cubic com- plexity in terms of the filter size. We study the infor- mation capacities of low-complexity iterative approxi- mations of these linear filters, and show that for large operation ranges, these simple iterative filters provide nearly the same capacity as original complex filters. Conditional correlation reception of multi-access signals is limited by mutual interference, hence joint detection method have been studied for a number of years to improve the spec- tral efficiency. The optimal detection (2) has an exponential complexity, and is feasible only for very small systems. As a result, linear filters (l, 3)are considered as the low-complexity alternatives. Linear pre-processing can be understood as conditioning the channel between a single transmitter and its designated receiver. These filters turn the noise and interference into a Gaussian noise source affecting the equivalent single-user channel. The information theoretic capacity is then calculated using Shannon capacity formula for the AWGN channel (4). The linear reception of multi-access signals bears the general form as