In this thesis, an improved genetic algorithm (GA) for multi-user detectors is investigated in synchronous and asynchronous DS-CDMA systems. Although the embedded GA and multistage detector (E-GA/MSD) can also improve the conventional GA performance for multiuser detection, the MSD incurs high computational consumption when the users are growing. To eliminate the MSD operation at each generation, an improved GA with adjustable parent size and enlarged sampling space is developed. The proposed GA with much less complexity can achieve almost the same BER performance as the E-GA/MSD. Simulation results demonstrate that the proposed GA is near to an optimum detector. Additionally, we apply the proposed GA to a spatial diversity reception assisted multiuser detector over uncorrelated flat Rayleigh fading channels in synchronous CDMA systems. As expected, computer simulations also show that the proposed GA can achieve the lower BER performance than conventional GA.