Design of low complexity multiuser detection using information theoretic criteria

In this paper, we propose a novel low complexity minimum mean-squared based multiuser detection (MMSE MUD), which employs partial but essential information of the interferences, in a multichannel TDMA system. The new approach begins with the determination of the effective channel length of the interferences based on the information theoretic criteria. We then truncate the channel taps according to the taps’ power using thus obtained effective channel length. The computational complexity of the resulting MUD is thereby reduced. Analytic studies are also carried out to provide more insights into the proposed approach. The furnished simulations show that the new MUD offers close biterror-rate (BER) performance as the full complexity one but with substantially reduced complexity. Radio spectrum is a rare and important resource in a cellular system. We intend to improve spectrum efficiency of a TDMA system with base station employing antennas to enhance performance. One way to maximize the spectrum efficiency is to lower the frequency reuse factor. This approach, however, results in stronger co-channel interferences (CCIs), which impairs the communication quality. The MUD technique [l] is an efficient process to mitigate the impairment caused by the CCIs. The optimal MUD algorithm, nevertheless, always calls for high complexity and thus suboptimal algorithms such as MMSE MUD are of more practical interest. To further facilitate real time implementations, several low complexity MUD schemes, which utilize only partial information of the users, have been addressed [l, 21. For example, the principal component method [l] retains only the eigenvectors corresponding to the largest eigenvalues in order to reduce the rank of the received data. Reference [2] considers a simple scheme of truncating the CC1 channel taps based on the taps’ *This work was supported by National Science Council of R.O.C. under contract 89-2213-E-011-106. power to reduce the complexity. All of these methods, however, need a prescribed truncation rate which is in general hard to determine. In this paper, we propose a new information based low complexity MUD by appropriately truncating the CCIs channel length. To retain the essential information of the CCIs in the truncation process without serious performance degradation, we first determine the effective channel length based on the information theoretic criteria, i.e. the Akaike’s information theoretic criteria (AIC) [3] and the minimum distance length (MDL). Based on thus obtained channel length, we then truncate the CC1 channel taps with weaker power. The computational complexity of the resulting MUD is thereby reduced. Meanwhile, since the truncated taps are redundant (account for negligible information) in the channels, they can thus be disregard and the effects of mismatch channel is negligible in the detection. Analytic studies are also carried out to provide further insights into the proposed approach. The furnished simulation results show that compared to the full complexity MUD, this new one can provide very close performance but with substantially reduced complexity. II. COMMUNICATION MODEL In the wireless communication model, we consider the scenario of m receiving antennas in a multi-channel and multiuser environment. In a TDMA system, the CCIs and the desired user transmit simultaneously in the same frequency band, after sampling, the received signal X can be expressed as XI mxnr) --Id d(mxlr) d(lrxnr) S