粒子濾波算法應用在多輸入多輸出天線正交分頻多工的訊號檢測之研究

Abstract

在多輸入輸出正交分頻多工的系統上，由於訊號之間的干擾，因此在接收端要做訊號的檢測的複雜度比單接收天線時的正交分頻多工系統來得複雜。特別是利用最大概似法(ML)作訊號檢測時，接收機的複雜度將會隨着天線的數量的增加或調變的不同呈現指數的增加。這篇論文主要是利用粒子濾波算法應用在多輸入多輸出正交分頻多工的訊號。粒子濾波算法利用統計的原理，造出相對應的事後機率用以作訊號檢測，來達到接近最大概似法的效能的同時減低複雜度。我們再使用一些方法合併粒子濾波算法去得到接近最大概似法的效果，在模擬顯示出粒子濾波演算法的結效能和我們提出的一些改良方法的效能比一般的VBLAST MMSE OSIC 更接近最大概似法的效能，而且其複雜度遠少於最大概似法。

In multiple-input multiple output orthogonal frequency division multiplexing (MIMO-OFDM) system, data detection become more complicated than single input single output (SISO) system especially for Maximum likelihood (ML) detection scheme. The complexity for ML detection scheme will increase exponentially as either the number of transmitting antennas or modulation order increases. In this thesis, we introduce the use of particle filtering to approximate a posteriori distribution so that we can use Maximum a posteriori (MAP) detection scheme to detect signals. We also present some new methods combined with particle filtering for data detection to mitigate the error propagation problem in either spatial multiplexing system or MIMO-OFDM system with space frequency block code system. These proposed methods have an improvement as compared with V-BLAST MMSE OSIC receiver in both systems. Simulations show that the BER performance for our proposed methods will approach to the ML decision algorithm as compared with VBLAST MMSE OSIC and the complexity is lower than ML decision algorithm.