Research On The Multi-user Detection Algorithm Based On The EM Method

Direct sequence code division multiple access (DS-CDMA) communication systems have very good anti-interference property and high channel utilization ratio, which make it be the main air interface technique adopted by the Third Generation Mobile Communication Systems. However, the multiple access interference (MAI) arises because of the incomplete orthogonality between different users' signature waveforms. A DS-CDMA system will be degraded in performance and limited in capacity because of the existence of MAI. Therefore, as an effective technique to mitigate the MAI, the multi-user detection technique has become a hotspot in the research of modern communication.In this paper, the main cause of the MAI is analyzed by exploring the model of DS-CDMA systems and the traditional matched filter algorithm. After explaining the fundamental theory of the multi-user detection algorithm, several typical multi-user detection algorithms, such as the optimum multi-user detection algorithm and the decorrelation detection algorithm, are introduced. Based on the above discussions, two kinds of joint detection algorithms are derived in detail, the performance of the two algorithms are also analyzed.The high computational complexity of the optimum multi-user algorithm prevents it from practical use. To solve this problem, the EM algorithm is applied to the field of multi-user detection. After explaining the EM algorithm, this paper gives the derivation of the multi-user detection algorithm based on the EM method. The multi-user detection algorithm based on the EM method is used to look for the maximum-likelihood estimation of users' data iteratively in a DS-CDMA system. It can obtain good detection performance and low computational complexity at the same time. But the traditional multi-user detection algorithm based on the EM method can't achieve ideal performance because of the existence of the "invalidated modification" problem under the constraint condition, which the algorithm should satisfy. This paper analyzes the reason of the problem mentioned. And also, a modified algorithm is proposed in this paper. In the modified algorithm, the discrete user data are supposed to have continuous value in the EM iterative processing, and the estimated value of each iteration cycle is limited, the limitation threshold is got by simulation. To get better BER performance, the estimate which has minimum absolute value is flipped and compared after the iterative processing. By this way, some kinds of errors, which are caused by the convergence to the zero value, can be corrected. The simulation result shows that the BER performance of the modified algorithm is improved in comparison with that of the traditional algorithm.