Research On Channel Estimation Methods Of MIMO-OFDM System Based On Neural Network

With the rapid development of the communication industry,the demand for communication speed is increasing,and the requirements for the effectiveness and reliability of communication are also increasing.In order to obtain better communication quality,MIMO-OFDM technology has attracted more and more attention.In MIMO-OFDM communication systems,the results and complexity of channel estimation have a great impact on the communication process.It means that a powerful method is required to finish channel estimation,which could extract the nonlinear relationship between the received signal and the transmitted signal,and accurately recover original signal.What's more,channel estimation of MIMO-OFDM communication system mainly faces the following challenges.Firstly,the signal will be affected by the uncertain noise during the transmission process,and the signal may have different degrees of fading,which make the nonlinear relationship between the received signal and the transmitted signal become more complicated.This will increase bit error rate and reduce communication reliability.Secondly,Since the channel estimation algorithm is implemented based on pilot sequences,and the pilot overhead is too large in a MIMO system,it will occupy a large amount of communication resources,resulting in reduced communication effectiveness.How to use fewer pilot sequences to complete channel estimation it is also a major challenge.Therefore,this thesis focuses on the above problems,and studying on the problem of communication reliability and effectiveness reduction caused by the channel estimation process of the MIMO-OFDM communication system.Focusing on the problem of inaccurate channel estimation in MIMO-OFDM communication system leads to the reduction of reliability,this chapter proposes a channel estimation method based on BP neural network.Firstly,using the pilot signal as training data to complete the training of the neural network.Secondly,taking use of the obtained neural network model to compensate and recover data signals.In addition,by adopting additional momentum and adaptive learning rate improvements to the neural network,the training time is shortened and the performance of the neural network model is improved.Finally,theperformance of method is verified and analyzed by combining with Matlab simulation software.The results reveal that when compared with the traditional method,this method not only reduces the bit error rate and improves the reliability of the communication system,but also does not involve the prior information of the channel during the implementation process.For the problem that the pilot overhead used for channel estimation in the MIMO-OFDM communication system is too large,which reduces the communication effectiveness.This chapter proposes a channel estimation method based on integrated neural network,which achieves channel estimation using fewer pilot signals while ensuring reliability.Firstly,combining with the idea of Bagging integration and using the self-sampling method to obtain multiple training subsets.The pilot sequence is used as the initial training data set and making the random sampling with putting back,which is to ensure that each training subset has enough samples.Multiple training subsets would be obtained when repeated many times.Secondly,using these subsets to train each neural network in the integrated neural network in parallel.And obtaining the integrated neural network model by combining diversity factor limits.Then,it can be used to recover the signal except the pilot sequence.Finally,the method is verified and analyzed on Matlab simulation software.The results indicate that when compared with the traditional algorithm,the method uses fewer pilot sequences under the same bit error rate.And the effectiveness of the communication system has been improved.What's more,the method has better stability than a single neural network model.