Blind Multiuser Detection For Uplink MC-CDMA

The multipath fading and multiple access interference (MAI) are the main factors affecting the performance of receiver in CDMA systems. The technique of OFDM can effectively suppress the intersymbol interference and resist the multipath fading, thus, the faded signals be received reliably. Therefore, the Multicarrier-CDMA technique, which brings together OFDM and CDMA, is a candidate for the 4^th generation mobile communication systems. Adopting multiuser detection (MUD) can reduce even eliminate the MAI, but it needs a training sequence, and a continuous sending of training sequence will occupy a lot of precious spectral resources. Blind detection, without needing training sequence, becomes a hot spot of research.Nowadays, blind detection is represented by the constant modulus algorithm(CMA), High-order moment , Signal subspace. Under a time variable multipath channel, a non-adaptive detection requires a computation of linear transform matrix for every symbol, which brings on high complexity. It will track the time-variable information parameters of the interrupted user's spread code by using the adaptive detector. According to the variation of the parameters, the MAI can be eliminated by iteration.This thesis introduces three structures of multicarrier-CDMA, and analyses the time-variable properties of multi-path channels and the MC-CDMA models under multipath channel. We improve the Affine Projection algorithm APA to adapt to blind multiusers detection. The improvement make use of the orthogonality between the signal subspace and noise subspace, and carry out the blind channel estimation by subspace separation. Track the variation of the information parameters of the desired user in time-variable channels and put the tracking results to the APA filters as parameters. During the update of the weight vector, the APA filter extracts the parameters of blind channel estimation, suppress the information from interfering users and recovers the information of desired user. We analyze, theoretically, the performance of the improved APA algorithm.The outstanding advantage of this algorithm is that it avoids the complex computation for inverse of the autocorrelation matrix. Besides, it has a simple structure and is easy to read and understand. We study the performance of the improved APAalgorithm under the multipath channel by computer simulation. The results indicate that, compared with the CMA algorithm, two algorithms have equivalent performance and robust under the same condition.