Indoor Dense Multipath Channel Model Identification And Parameter Estimation In Uwb Communication System

Because of its advantages such as high data rate,low power spectral density,short time pulse and large bandwidth,UWB communication system is widely used in many fields such as people's livelihood and military affairs.Because of its wide bandwidth,at the same time,the indoor environment is not only narrow but also has many obstacles,the ultra-broadband signal will show dense multipath phenomenon,which makes it difficult to identify and estimate the channel model.With the rise of machine learning and artificial intelligence revolution,artificial intelligence is more and more widely used in wireless communications.This paper presents the channel model identification and SNR estimation method based on neural network,and used the identified channel model and estimated SNR as prior knowledge,improve the channel estimation algorithm based on compressed sensing theory.In this paper,there are three main contents in the field of indoor dense multipath channel of UWB,which are channel model identification,SNR estimation and channel estimation.Firstly,this paper uses traditional channel characteristic parameter to identify channel models by support vector machine(SVM),the result of identification is very low.So this paper presents the channel model identification based on convolutional neural network method,the method based on IEEE802.15.4a channel model,taking advantage of quadrature pulses constructed by Gegenbauer polynomials as transmit signals,and generate the signals as the training and testing collection.A nalyzed the structure and parameters should be selected when channel model identification is realized in the convolutional neural network.To train and test the built network,got a classification accuracy which is pretty good.Secondly,this paper presents the SNR estimation method based on neural network.Analyzes and compares the performance of the convolutional neural network and the recurrent neural network in terms of SNR estimation.The result shows that the performance of the recurrent neural network is better than that of the convolution neural network in the estimation of time series signals.In this part,it also used the identified channel model as a priori knowledge to test the accuracy of the signal-to-noise ratio estimation in the case of whether the channel model is known,and it is proved that the known channel model helps improve the accuracy of the SNR estimation.Thirdly,this paper improves the method to estimate the parameters of channel by compressed sensing theory,and by using the known channel model and the signal-to-noise ratio.Number of guide frequency is analyzed at different channel model under different signal-to-noise ratio to estimate the influence of the error,implements the parameter adaptive channel estimation method,improve the efficiency of the channel parameter estimation.The results of channel parameter estimation are applied to the des ign of Rake receiver,and the low bit error rate is obtained.The channel model identification and paremeter estimation method used in Ultra-wideband communication system of indoor dense multipath channel which is present in this paper,based on neural network and compressed sensing algorithm respectively,using the obtained conclusions as prior knowledge step by step,compared with traditional methods,it reduced the artificial intervention factors,and better results have been achieved.