Hybrid MMSE and SIC for Multiuser Detection

• Main Authors: Chia -Hung, Tsai, 蔡佳宏
• Other Authors: 方文賢
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/97084058668270918427

碩士 === 國立臺灣科技大學 === 電子工程系 === 89 === Near-far problem limits both of the system performance and the overall network capacity in the reverse link of a DS-CDMA system. For this, various approaches have been addressed, to overcome these difficulties, among which the multiuser detection (MUD) has been proved to be effective. Multi-user detection (MUD) has demonstrated to be an effective approach to resolve this problem. The optimal MUD, however, in general calls for substantial computational overhead. As such, some suboptimal ones, such as minimum mean-squared error (MMSE) MUD and successive interference cancellation (SIC) MUD, are of more practical interests. Nevertheless the computational complexity of the MMSE MUD is still high, while the SIC MUD suffers the latency problem for the detection of the weakest signals. In light of this, this thesis proposes a hybrid MMSE and SIC MUD which possesses the advantage of both detection schemes, but with alleviated aforementioned drawbacks. The approach first divides the users into several groups with each group consisting of users with a close power level. The SIC MUD is then used to distinguish users among different groups, while the MMSE MUD is used to detect users within each group. As such, the computational complexity of the MMSE MUD and the latency problem in the SIC MUD can thus be greatly mitigated. Besides, since the SIC MUD substrates the reconstructed signal from the original one stage by stage, the reconstructed error can then propagate and eventually may destroy to detection process. To overcome this, this thesis also addresses a new dynamic mapping function following the MMSE MUD carried out in each group of the proposes hybrid approach. This new mapping function can not only alleviate the false alarm induced in the reconstruction process, but also possesses dynamic adjusting scheme according to the signal characteristic. Furnished simulations show that the new approach outperforms both of the traditional MMSE MUD and SIC MUD algorithms.