Adaptive wavelet neural networks for signal detection in DS-CDMA system

The Multiple access interference (MAI) is the major factor that limits the performance and capacity of a nonorthogonal Direct sequence Code division multiple access (DS-CDMA) system. By using the adaptability of highly parallel structure neural networks and the excellent approximation ability of wavelets, two kinds of Adaptive wavelets neural networks (AWNN) signal detectors are proposed in the paper, in which the inputs of detectors are the received signal vector corresponding to a single interesting user sampled at the chip rate, named by AWNN single-user detector, respectively, and to all or partial active users sampled at the bit rate after passing through a matched filter, named by AWNN multiuser detector and partial users AWNN multiuser detector. The complexity of the multiuser detectors only depends on that of wavelet networks. The performance analysis of the proposed detectors compared with the matched filters under single-user and multiuser systems and the multiuser detector based on multilayer perceptrons are carried out by Monte Carlo simulations. Results show that the adaptive wavelet neural networks multiuser detectors are superior to other detectors mentioned above.