The Application Of Noise Independent Component Analysis In Multi-user Detection

Code Division Multiple Access is self-intcrfered system, where multi-accessinterference (MAI) and Near-far effect arc existing as the main factors restricting CDMA system capacity. These years, be considered as one of the key techniques forthe 3rd Generation (3G) mobile communication system, Multiple User Detection(MUD) for CDMA system made its obvious progress. MUD reduces the multi-accessinterference and solves the Near-far problem and improves system capacity. The blindadaptation schemes do not require training sequences and the interference user'sinformation. It only needs the prior knowledge of the signature waveform and thetiming of the desired user. The blind adaptation schemes are the promising one withacceptable implementation complexity, while simultaneously eliminating the need ofa plenty of priori system information and can be implemented as a blind MUD.The multi-user detection based independent component analysis belongs toblind MUD. The content of this paper is about this kind of blind MUD. In this paper,we first introduce the basis concept and theory on the multi-user detection and somemostly important multi-user detection methods, then summarize the development ofblind MUD. We also expatiate on theory of independent component analysis (ICA)and the FastICA based Negentropy and Newton's method, The ICA model ignore thenoise so the ICA algorithm will have a bias. In order to remove the bias we introducea noise ICA based the noise model and deduce the multi-user detection method basedthis algorithm. Before the noise algorithm, we implement dimension reduction toapprove the iteration availability. To compare the performance of the two kinds ofmulti-user detection methods which are based the normal ICA algorithm and the noisealgorithm, we simulate the two methods. Simulation results indicate that noise ICAalgorithm method has better performance especially in the case of lower SNR, moreusers and little symbols with the iteration number as much as normal ICA method.