在正交分頻多工通訊系統中,部分的頻寬會犧牲給訓練符號(training symbols)作為通道估計之用。為了減少訊練序列的長度同時提升頻寬的使用率,我們利用決策反饋機制(Decision-Feedback)改良的卡爾曼濾波(Kalman Filter)時域通道響應追蹤估測器。在卡爾曼濾波估測器中,增加使用調適性變異數估測(adaptive variance estimation)及以提升估測器的精確性,以錯誤檢測器來偵測決策錯誤,並提出補償方法來避免決策錯誤的蔓延(Decision-error propagation)。在實驗模擬中,我們比較所提出的檢測器與補償方法在不同的通道模型中的性能效果,用以呈現所提方法之效能。 In an orthogonal frequency division multiplexing (OFDM) system, part of bandwidth is occupied by training symbols for channel estimation. In order to reduce the period of the training sequence while improving the bandwidth efficiency, we study a new time-domain channel tracking algorithm using a Kaman filter in the modified decision-feedback loop. In company with adaptive variance estimation in the Kalman filter for enhancing channel estimation, some decision error detectors and compensation methods are proposed for avoiding decision error propagation to reduce the error probability. In simulation results, we compare the performances of the proposed decision error detectors and compensation methods for different channel models to show the efficiency of the proposed algorithm