Ultra-Wideband Channel Estimation Based On Compressed Sensing And Deep Learning

Under the framework of Shannon theorem,it is a main trend to boost the channel capacity by increasing channel bandwidth in future communications.Ultra-wideband(UWB)technology transmits information via extremely wide bandwidth channels.Therefore,the study of key issues of ultra-wideband technology such as channel estimation plays a significant role in the development of future communication systems.There are two technical challenges for UWB channel estimation:1.Because UWB systems occupy extremely wide spectrum,the receiver needs very high Nyquist sampling frequency,which will increase the hardware cost greatly.Fortunately,applying compressed sensing(CS)to UWB systems can reduce the sampling frequency effectively.However,the application of CS brings the problem of noise folding and thus affects the accuracy of channel estimation.2.UWB channel is complicated and time-varying,and different channel types differ from each other greatly.Thus,UWB channel classification is a key issue for adaptive transmission in dynamic environment.This thesis studies the CS-based UWB channel estimation and deep-learning-based UWB channel classification.The purpose of this work lies in two folds.One is to improve the accuracy of channel estimation by suppressing noise folding of compressed sensing.The other is to improve the adaptability of UWB systems in time-varying environment by using deep-learning-based channel classification.Our contribution is as follows:1.In order to reduce the effect of noise folding in UWB compressed sensing,an channel estimation scheme capable of suppressing noise is proposed.Based on the power distribution statistics of UWB signal and additive Gaussian noise on eigen-based dictionary,truncated measurement is proved to be feasible for noise suppression.By optimizing the truncated measurement,an adaptive measurement method is proposed so as to maximize the signal to interference plus noise ratio of the output signal.In this way,the noise folding is suppressed effectively.Simulations have been carried out to verify the performance of the proposed in terms of channel correlation coefficient and bit error rate of coherent receiver.The simulation results suggest that the proposed scheme can be used to improve the channel estimation precision effectively.2.In order to improve the adaptability of UWB systems in time-varying environment,a channel classification method based on deep learning is proposed.The proposed identifies the current channel type using a predesigned and trained neural networks in the first stage,where the noisy measurements obtained by CS are taken as input.Then,an ensemble decision process is conducted to improve the classification accuracy further.Simulation results show that the proposed method can classify UWB channels accurately.Based on the channel classification result,an adaptive transmission method is designed to improve the system throughput.Simulations have been carried out to compare the throughput of different transmission schemes under several channel conditions.Simulation results confirm that the proposed adaptive transmission scheme can improve the system throughput effectively in dynamic environments.