Research On OFDM Signal Detection Algorithm Based On Deep Learning

This article mainly studies how to improve the signal detection capability of OFDM wireless communication systems by using deep learning algorithms.Traditional signal detection algorithms have poor performance when the number of pilots in the OFDM wireless communication system is small and there is no cyclic prefix.Although the OFDM signal detection algorithm of deep neural network improves the signal detection performance of the OFDM system to a certain extent,the performance improvement is small.This paper proposes to apply Long Short-Term Memory(LSTM)to OFDM signal detection to improve the signal detection capability of the OFDM system.On this basis,it proposes a neural network based on Bi-directional Long Short-Term Memory(Bi-directional Long Short-Term Memory).Term Memory,Bi LSTM)signal detection network.First,aiming at the situation where the DNN signal detection network is not ideal for OFDM system signal detection,this paper proposes the LSTM signal detection network and introduces the framework structure and training process.The LSTM signal detection network mainly includes the input layer,the LSTM layer,and the fully connected layer.,Softmax layer and classification layer,the LSTM signal detection network can extract information between the input data,so it has better performance than the DNN network.The network needs to be trained on the data generated under the existing channel model,and then Put the trained network into the OFDM system for application.Simulation shows that the LSTM signal detection network has a higher performance improvement compared with the traditional algorithm in the case of different number of pilots and with or without cyclic prefix.Compared with the DNN signal detection network,when the symbol error rate is 10^-2,With a 1d B performance improvement.Secondly,in view of the slow convergence speed of the previously proposed LSTM signal detection network,this paper designs a signal detection network based on a two-way long and short memory neural network.The network mainly includes:input layer,Bi LSTM layer,fully connected layer,softmax layer and classification layer.First,you need to build a Bi LSTM network,and then use the data generated under the 3GPP channel model to train the built neural network,and finally you can apply the trained neural network to the OFDM system to detect the signal of the entire system.The simulation shows that compared with the traditional signal detection method,the Bi LSTM signal detection network has a 5?6d B performance improvement when the symbol error rate is 10^-3.Compared with the same type of algorithm using end-to-end LSTM signal detection network,this algorithm has a performance improvement of 1B when the symbol error rate is 10^-3.The Bi LSTM signal detection network has a faster convergence speed and a higher loss function value.