Academic
Publications
A Frequency Domain State-Space Approach to LS Estimation and Its Application in Turbo Equalization

A Frequency Domain State-Space Approach to LS Estimation and Its Application in Turbo Equalization,10.1109/TSP.2011.2135857,IEEE Transactions on Signa

A Frequency Domain State-Space Approach to LS Estimation and Its Application in Turbo Equalization   (Citations: 1)
BibTex | RIS | RefWorks Download
A frequency domain state-space approach to the least squares (LS) estimation for finite impulse response (FIR) system identification is proposed by exploiting the Toeplitz structure of the data matrix, leading to two low-complexity FFT (fast Fourier transform)-based unbiased recursive estimators. The proposed re- cursive estimators are applied to the smoothing estimation of time- varyingfrequencyselectivechannelsinturboequalizationsystems, and channel estimation approaches with complexity per symbol are developed, where is the number of channel taps. IndexTerms—Channelestimation,factorgraphs,Gaussianmes- sagepassing,least squares,state-space approach, time-varyingfre- quency selective channels, turbo equalization.
Journal: IEEE Transactions on Signal Processing - TSP , vol. 59, no. 7, pp. 3288-3300, 2011
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...1) By carrying out channel estimation and detection jointly, the proposed low-complexity EM based approach can significantly outperform alternative approaches (with similar complexity) where channel estimation and detection are implemented separately, e.g., the approaches in [17]–[19] (in which the uncertainty due to the estimate of the symbols is accommodated into the additive noise for channel estimation, and vice versa for symbol ...
    • ...The length of the frequency domain channel response in model (6) is , but the actual degree of freedom of the channel is (since the length of is ). This means that there are some constraints among the entries of , which should be properly exploited [19]...
    • ...As in [19], to reduce the complexity, the following approximation is adopted in the forward covariance matrix updating for the multiplication node...
    • ...In the approaches [17]–[19] where the channel estimation and detection are implemented separately, the complexity of detection is and that of channel estimation is (in [17] and [18]) or (in [19])...
    • ...In the approaches [17]–[19] where the channel estimation and detection are implemented separately, the complexity of detection is and that of channel estimation is (in [17] and [18]) or (in [19])...
    • ...The complexity of the proposed approach depends on ,b ut can be a small number, e.g., in the simulations shown in Section IV, the maximum of is only 5. The proposed approach can achieve significant performance improvement by using the EM iteration compared with the approaches in [17]–[19], which will be demonstrated in Section IV...
    • ...As in [9] and [17]–[19], with the approximation that the channel is static within a short time duration, we partition the length- signal...
    • ...The MSIE performance and the BER performance of the proposed approach, the approach in [17], the approach in [18] (with approximate rule 1), and the approach in [19] (with CE2) are shown in Fig. 3. It can...

    Qinghua Guoet al. EM-Based Joint Channel Estimation and Detection for Frequency Selectiv...

Sort by: