In Space Division Multiple Access - Orthogonal Frequency Division Multiplexing (SDMA - OFDM) wireless communication system, the classical Minimum Mean Square Error (MMSE) and Maximum Likelihood (ML) Multiuser Detection (MUD) techniques suffer with poor performance and high complexity. The Minimum Bit Error Rate (MBER) using Genetic Algorithm (GA) has become a good alternative to these techniques by performing better than MMSE with less complexity. However, all these above mentioned techniques require additional channel estimation block and also are restricted to sub-optimal performance. In this research contribution, a multiuser detection scheme using complex Neural Network (NN) structure such as complex Multi Layered Perceptron (MLP) model is proposed for this system. It has been found through simulation studies that the proposed MLP aided MUD at receiver's end performs better than those classical MUD techniques in terms of BER performance and complexity. © 2012 American Institute of Physics.