Minimum bit-error-rate linear multiuser detector for time-varying and frequency-selective fading DS-CDMA

The capacity of the Direct Sequence ? Code Division Multiple Access (DS-CDMA) systems is limited by multiple access interference (MAI). Nowadays, the matched filters (MF) are used in the DS-CDMA systems based on IS-95 standard. Therefore it has negative effect on the number of simultaneous users of the systems. The BER performance of the system decreases while the number of active users is increasing. A minimum bit-error-rate (MBER) multiuser detector (MUD) is considered for DS-CDMA communication systems with time-varying and frequency-selective fading channels. The bit-error-rate (BER) cost function of the proposed MBER MUD is highly non-linear and may have several local minimum. It is shown that with some appropriate constraints the BER cost function of the MBER MUD is equivalent to a constrained optimization problem which has a unique global minimum in the feasible region. An efficient Newton method with a barrier parameter is developed for finding the coefficients of the proposed MBER MUD. The BER performance of the MBER MUD is compared to decorrelating (DEC) detector, linear minimum mean-square error (LMMSE) detector, and the maximum likelihood detector for time varying and Rayleigh distributed frequency-selective fading DS-CDMA channels. Monte-Carlo simulations show that the BER of the MBER MUD can be significantly lower than that of the DEC and the LMMSE multiuser detectors. Proposed MBER algorithm is used for; Time-invariant and frequency-selective fading channels (BER), Time-varying and frequency-selective fading channels (average BER), channel models. Then, the BER cost function is converted into a constrained optimization problem. The MBER MUD is obtained by solving the constrained optimization problem. The proposed algorithm minimizes the BER cost function directly subject to the constraint that is a convex set of the feasible region. The Newton's method with a barrier parameter is used to find the filter coefficients of the proposed MBER MUD, which has lower computational complexity than the ML detector. The analysis and simulations results confirm that the proposed MBER MUD minimizes the BER cost function and can achieve lower BER than the DEC and the LMMSE detectors in medium or high SNR cases. Furthermore, the MBER MUD has a lower complexity than the maximum-likelihood (ML) multiuser detector but achieves approximately similar BER performance.