本論文主要是探討在分碼多工系統中,不同機率初始值設定對於機率資料連結演算法之多使用者偵測器效能之影響。雖然使用機率資料連結演算法之多使用者偵測器可以得到近乎最佳化多使用者偵測器的效能,但其計算複雜度相較於最小均方誤差多使用者偵測器仍然偏高。為了增快其收斂速度,雖有人提出解相關多使用者偵測器之輸出訊號之軟性決定作為機率初始值之設定,此一方法,雖可加快其收斂速度,但在平坦衰落通道之MC-CDMA系統中,當系統中人數過多時,會產生位元錯誤率提高的問題。為解決此一問題,本論文將更進一步的探討使用多種次佳化多使用者偵測器之輸出訊號之軟性決定來設定機率初始值對於機率資料連結多使用者偵測器之影響,藉由電腦模擬的方式以驗證其可行性與效能分析。同時也將針對於平坦衰落通道MC-CDMA系統中人數增加所產生位元錯誤率提高的問題,解析其原因並加以提出改善的方式,以其在維持相同位元錯誤率下,加快其收斂速度。此外,本論文亦將探討機率初始值的設定對於使用者有無依有效能量排序及遠近效應對機率資料連結多使用者偵測器的影響。
The main purpose of this thesis is to research into the performance of multi-user detector using probabilistic data association (PDA-MUD) by setting different initial probabilities in CDMA systems. The performance of PDA-MUD is close to the optimal multi-user detector, but computational complexity of PDA-MUD is higher than the linear minimum mean-square error multi-user detector. To reduce the computational complexity, a modified PDA-MUD was proposed by using the initial probabilities setting related to the soft decision of the output signals in decorrelating multi-user detectors. But this method will increase the bit error rate in some cases. To solve this problem, we widely investigate the performances of PDA-MUD by using different initial settings based on soft decisions of the outputs of various suboptimal multi-user detectors. Finally, we propose a new modified PDA-MUD scheme that can improve the convergence speed while maintaining the same bit error rate as the original one in these cases. In addition, this thesis also analyzes the performance effects of users sorting and near-far effect on the PDA-MUD with different initial probabilities settings.