Research Of Channel Intelligent Estimation Technology For 3D MIMO Channel

Three-dimensional Multiple Input Multiple Output(3D MIMO)technology enhances the capacity of system by extending the vertical dimension of antenna based on Multiple Input Multiple Output(MIMO)technology.MeanWhile,Orthogonal Frequency Division Multiplexing(OFDM)technology can improve the communication system's frequency band resource utilization and multipath fading resistance.the 3D MIMO system that using OFDM technology has become an important part of next-generation communication technology,so research related to 3D MIMO technology plays a decisive role in practical engineering applications.3D MIMO technology extends the vertical dimension of the channel,making the wireless channel more complicated than other situations.Studying channel estimation methods with low pilot resource occupancy,low complexity and high accuracy is essential for improving the communication quality of 3D MIMO systems.This research mainly studies the channel estimation algorithm in 3D MIMO channel.This research mainly works as follows:Firstly,it is verified by simulation that 3D MIMO channel is sparse in the time domain,and the compressed sensing technology can be used to get channel information.Then,considering that the existing compressed sensing reconstruction algorithm does not consider the noise effect,and the channel is susceptible to noise interference,an improved MRAMP algorithm is proposed.The algorithm uses the ratio of signal residual as the iterative termination condition,so,it can stop iteration timely under various SNR environments,avoiding the over-estimation problem caused by excessive noise energy.The simulation results show that the compressed sensing channel estimation technology can greatly reduce the number of pilots and save band resources compared to traditional Least Square(LS)algorithm.And compared with Sparsity Adaptive Matching Pursuit(SAMP)algorithm,the MRAMP algorithm can reduce the mean square error by an order of magnitude,which improves the accuracy of the compressed sensing channel estimation.Then,considering that the compressed sensing reconstruction algorithm has high complexity and slow convergence,machine learning technology can realize the mapping of complex relationships between input and output through a specific model,a low complexity channel intelligent estimation method based on Convolution Neural Network(CNN)is proposed.The method converts the channel estimation problem into a super-resolution problem by considering the channel frequency domain response of one OFDM block as a two-dimensional image.By using the method of machine learning,a three-layer convolution network CECNN can recover the high-resolution image from a low-resolution image.the high-resolution image represents the complete channel response and the low-resolution image represents the channel response of pilot.the results of simulaton show that the channel estimation mean square error of CECNN model is smaller than LS algorithm,and CECNN model is an effective channel estimation method.Finally,with the Simulink platform,a 3D MIMO system based on the WINNER? urban macrocell scenario is built.The different system performance changes is verified when using LS,MRAMP algorithm or CECNN model for channel estimation.The results show that the mean square error of channel estimation of MRAMP algorithm and CECNN model is smaller than LS algorithm.MRAMP algorithm and CECNN model can reduce bit error rate of the system and improve system performance.Compared with MRAMP algorithm,the CECNN model greatly reduces the time complexity of channel estimation,and the channel estimation mean square error and system bit error rate are only slightly higher than MRAMP algorithm.