eduzhai > Applied Sciences > Engineering >

Beamspace Channel Estimation in Terahertz Communications A Model-Driven Unsupervised Learning Approach

  • Save

... pages left unread,continue reading

Document pages: 26 pages

Abstract: Terahertz (THz)-band communications have been one of the promisingtechnologies for future wireless networks that integrate a wide range ofdata-demanding applications. To compensate for the large channel attenuation inTHz band and avoid high hardware cost, a lens-based beamspace massivemultiple-input multiple-output (MIMO) system is considered. However, the beamsquint effect appeared in wideband THz systems, making channel estimation verychallenging, especially when the receiver is equipped with a limited number ofradio-frequency (RF) chains. Furthermore, the real channel data cannot beobtained before the THz system is used in a new environment, which makes itimpossible to train a deep learning (DL)-based channel estimator using realdata set beforehand. To solve the problem, we propose a model-drivenunsupervised learning network, named learned denoising-based generalizedexpectation consistent (LDGEC) signal recovery network. By utilizing the Steinsunbiased risk estimator loss, the LDGEC network can be trained only withlimited measurements corresponding to the pilot symbols, instead of the realchannel data. Even if designed for unsupervised learning, the LDGEC network canbe supervisingly trained with the real channel via the denoiser-by-denoiserway. The numerical results demonstrate that the LDGEC-based channel estimatorsignificantly outperforms state-of-the-art compressive sensing-based algorithmswhen the receiver is equipped with a small number of RF chains andlow-resolution ADCs.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×