Research On Multiuser Detection Technology In MC-DS-CDMA System

Now communications are moving rapidly toward the fifth generation of UWB communications systems.As a communication technology with high spectrum utilization,MC-DS-CDMA has good anti-interference,anti-fading and low detection probability.It has high research value and becomes one of the core technologies in future mobile communications.However,as the number of users increases,since the spreading codes are not completely orthogonal,and each sub-carrier is affected by the channel differently,the auto-correlation characteristics and cross-correlation characteristics of signals of various users in the system are destroyed,resulting in serious multiple access interference.In order to solve this problem,this paper studies the multi-user detection technology.The main work is as follows:Firstly,the CDMA and OFDM systems are studied,and the block diagram of their transceivers is given.Then the MC-DS-CDMA system based on the CDMA and OFDM system and its transceiver model are studied.Then the traditional detection which treats other users' interference signals as noise and directly makes decisions after demodulation and dispreading is described.But this detection's effect is very poor.Secondly,several existing multi-user detectors are studied.Then,a focus is placed on multi-user detectors based on BP neural networks.This detector utilizes the powerful classification and identification capabilities of BP neural networks to detect multi-user signals.Compared with other multi-user detectors,the BER of the user signal after detection is even lower.Then a multi-user detector based on simulated annealing neural network is proposed.Compared with multi-user detector based on BP neural network,this detector uses simulated annealing algorithm to train BP neural network and judges before updating the weights and thresholds.If the prediction results obtained during the current iteration are better than those obtained during the previous iteration,update the weights and thresholds.Otherwise,according to a certain The probability determines whether to update the weights and thresholds during the current iteration.Afterwards,the simulation proves that the detector in MC-DS-CDMA not only has better BER performance,but also has less complex than multi-user detection based on BP neural network.Finally,in order to further reduce the multiple access interference in the user signal,the interference cancellation algorithm is studied and analyzed.Combined with the parallel interference cancellation algorithm,a parallel interference canceller based on the simulated annealing neural network is proposed.The detector is first based on the use of The simulated annealing neural network multi-user detector completes the preliminary detection of each user signal,then estimates the multiple access interference in each user,and subtracts the estimated multiple access interference from each user's signal,and then completes the user through the decision.Signal detection,and finally through simulation experiments show that in the MC-DS-CDMA system,parallel interference canceller based on simulated annealing neural network and BP neural network-based multi-user detector improve the bit error rate of about 1d B Performance,and have a good ability to resist near-far effects.