A Research Of Polar Codes Decoding Based On Deep Learning

In the future era of all-things interconnection,there is a demand for massive machine-type communication applications.These applications are characterized by a small amount of data,but have strict requirements on delay and reliability.Therefore,the short length and high-performance channel coding schemes are urgently needed.Polar code is an emerging channel coding technology with outstanding performance in recent years.A Polar code scheme proposed by Huawei has been approved by3 GPP as a 5G control channel error correction coding scheme.On the other hand,due to its excellent performance in the field of image recognition and natural language processing,deep neural networks are permeating more disciplines and have achieved encouraging results.It has important theoretical significance to study the decoding algorithm of Polar codes and introduce the deep learning technology into Polar codes decoding scheme.In this paper,the principle of channel polarization and the encoding and decoding scheme of Polar codes are described and the main Polar codes algorithms and their drawbacks are analyzed.A novel joint demodulation-decoding scheme for multi-antenna Polar Codes is proposed.A Convolutional Neural Network(CNN)is constructed,it contains six learned layers including four convolutional,two fully-connected and one output layer.The mean squared error(MSE)loss function is adopted and polar-coded multi-antenna dataset is generated to train the network,the features between the bits of Polar Codes can be extracted and a well-fitted Polar Codes decoding function is achieved.Simulation results show that the proposed scheme has superior performance in BER over the existed fully connected network based scheme.Moreover,in all the simulations with different code rates,the proposal achieves better results than the original one,which proves superior generalization of the joint demodulation-decoding scheme for multi-antenna Polar Codes.