eduzhai > Applied Sciences > Engineering >

Model-Driven DNN Decoder for Turbo Codes Design Simulation and Experimental Results

  • Save

... pages left unread,continue reading

Document pages: 31 pages

Abstract: This paper presents a novel model-driven deep learning (DL) architecture,called TurboNet, for turbo decoding that integrates DL into the traditionalmax-log-maximum a posteriori (MAP) algorithm. The TurboNet inherits thesuperiority of the max-log-MAP algorithm and DL tools and thus presentsexcellent error-correction capability with low training cost. To design theTurboNet, the original iterative structure is unfolded as deep neural network(DNN) decoding units, where trainable weights are introduced to the max-log-MAPalgorithm and optimized through supervised learning. To efficiently train theTurboNet, a loss function is carefully designed to prevent tricky gradientvanishing issue. To further reduce the computational complexity and trainingcost of the TurboNet, we can prune it into TurboNet+. Compared with theexisting black-box DL approaches, the TurboNet+ has considerable advantage incomputational complexity and is conducive to significantly reducing thedecoding overhead. Furthermore, we also present a simple training strategy toaddress the overfitting issue, which enable efficient training of the proposedTurboNet+. Simulation results demonstrate TurboNet+ s superiority inerror-correction ability, signal-to-noise ratio generalization, andcomputational overhead. In addition, an experimental system is established foran over-the-air (OTA) test with the help of a 5G rapid prototyping system anddemonstrates TurboNet s strong learning ability and great robustness to variousscenarios.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×