Blind Mutiuser Detection Based On Improved Subspace Tracking Algorithm

The CDMA is the key technologies of third generation mobile communications. As to achieve multi-user shared channel approach, it assign different spreading code to different users. Compared to FDMA and TDMA systems, the CDMA system has many advantages such as high bandwidth efficiency, soft capacity, soft switching, low power, security and so on. However, CDMA systems are still some shortcomings, the main interferences are the I SI and multiple access interference. The ISI is compensated through balanced technology. The MAI was produced by multiple users sharing a channel, while spreading code of different users is not completely orthogonal. The primary method of inhibiting multiple access interference is called multi-user detection.Blind multi-user detection only needs observations, without getting the training sequence.Because of needing less data, we can make full use of bandwidth resources. Blind multi-user detection based on subspace divided the received signal autocorrelation matrix into signal subspace and noise subspace. The data of signal subspace are only used in the detection, the data of noise subspace can be ignored.These will simplify the blind multi-user detection.The paper improved the subspace-based blind multi-user detection algorithm from the correlation of the signal and the forgetting factor, the main work is as follows:Firstly, this paper analyzes the basic principles, performance measurement and classification of multi-user detection, research status and current on blind multiuser detection based on subspace algorithms in the world, and simulates the capability of several typical multi-user detection algorithm.Secondly, via studying the subspace-based iterative algorithm based on optimization theory, we found the cost function only considers information before the current time without taking into information after the current time. But BELLO tells us that time-varying channel correlation can be showed by coherent time, and the channel is only related in the coherence time. So we design a subspace tracking algorithm based on double-side cost function by adding information after the current time in the cost function.The simulation indicates the improved algorithm's BER is smaller than the Original.Thirdly, via studying PASTd algorithm, we found that its the tracking error and the convergence rate are due to the forgetting factors. For smaller forgetting factor, the tracking error is bigger, and the convergence rate is faster; For larger forgetting factor, the tracking error is smaller, and its convergence rate is slower. When forgetting factor will be revised as the function of the tracking error, it can decrease the tracking error, but also to speed up its convergence. When the channel is mutated, forgetting factor can be adjusted by the tracking error in time. The Matlab simulation shows that the convergence error and rate of the variable-forgetting-factor PASTd are between the two PASTd algorithms for the forgetting factorβ=0.995 andβ=0.999.