由於無線電資源有限，分時雙工架構下的巨量多輸入多輸出系統(mMIMO)通常需採用非正交領航序列，進而導致領航序列污染問題，嚴重影響通道估計的準確性。雖然最小均方誤差(MMSE)估計法具備良好效能，但通常仰賴事先已知的通道共變數矩陣，傳統解法多需透過多細胞協作與額外無線資源取得相關統計資訊。 本研究提出一種基於多步驟卡爾曼濾波器(Kalman Filter, KF)的通道估計方法，無需跨細胞合作或額外資源支持。該方法以領航與數據輔助模式交錯運行，透過迭代方式進行估計，不僅能有效近似最小均方誤差效果，還能解決通道估計的時效性問題。通道的共變數矩陣則以樣本共變數動態估計，分別對目標使用者與同頻干擾使用者進行建模。 模擬結果驗證了該方法的有效性，在無需額外計算或通道統計資訊的情況下，即可成功抑制領航污染，其估計準確度接近於所有共變數已知的理想情況。;Due to the limited availability of radio resources, massive multiple-input multiple-output (mMIMO) systems operating under a time-division duplex (TDD) framework often require the use of non-orthogonal pilot sequences, which leads to pilot contamination and significantly degrades the accuracy of channel estimation. Although the minimum mean-square error (MMSE) estimation method provides superior performance, it typically relies on prior knowledge of channel covariance matrices. Traditional approaches usually require multicell cooperation and additional radio resources to obtain the necessary statistical information.This study proposes a novel channel estimation method based on multi-step Kalman filtering (KF), which operates without the need for intercell coordination or extra resources. By alternating between pilot and data-aided modes in an iterative manner, the proposed method not only approximates the performance of MMSE estimation but also addresses the issue of outdated channel estimates. The channel covariance matrices are dynamically estimated through sample covariance matrices, capturing both the desired user and cochannel interferers.Simulation results confirm the effectiveness of the proposed approach, demonstrating that it can successfully mitigate pilot contamination and achieve estimation accuracy close to the ideal case where all covariance matrices are perfectly known—all without additional computation or statistical information.
