Research Of CDMA Multi-user Detection Based On Kernel-ICA

CDMA has become the mainstream technology of the third generation mobile telecommunication.CDMA system distributes orthogonal(or as orthogonal as possible) code to different users in order to realize multiple users to access the system at the same time and frequency. Due to random access and the correlation of spreading code, i.e. complete orthogonal between multiple users, it will cause the multiple access interference(MAI) in the end when multiple users are communicating. In order to get rid of the multiple access interference, Multiple User Detection (MUD) plays an important role.At first, this dissertation introduced the principle of multi-user detection and some sub-optimal multi-user detectors, and then summarized advantages and shortcomings and practical applications of MUD. The blind multiuser detection does not need any prior knowledge of the channel and the training sequence, so it is widely used in the mobile communication field.Secondly, this dissertation studied ICA (Independent Component Analysis) algorithm, the purpose of which is to separate the independent components from the observed linear mixed independent signals from multiple channel measuring. ICA is a new blind signal separation (BSS) technology, which is developing rapidly in recent years. Because of ICA without prior knowledge of the system's parameters, the estimation is based purely on the high order statistical properties of the signal.At last, this dissertation applied Kernel-ICA to MUD in CDMA. The conventional ICA-based MUD does not utilize the information of the spread sequences of users, but the expected user's spread sequence is known at the received end and has an ability in dividing signal. The MUD of conventional ICA can not make use of information of users as an result of wastes in the known resource. In this dissertation, Kernel Independent Component Analysis is that the signal from low-dimensional space is mapped to the high-dimensional space, in high-dimensional space, the target user spreading codes as a solution of mixed initial value of matrix. We solve the indeterminacy problem in ICA algorithms with spreading codes in high-dimensional space of Kernel ICA and this methd lows BER with good real-time. Simulation results indicate that the proposed algorithm is able to achieve better BER performance than the conventional matched filtering and conventional ICA approach in the same environment.