MIMO-OFDM system channel estimation method based on deep neural network

Abstract

The invention discloses an MIMO-OFDM (Multiple Input Multiple Output-Orthogonal Frequency Division Multiplexing) system channel estimation method based on a deep neural network. The method comprises the following steps: firstly, establishing an MIMO-OFDM system model based on a channel estimation method of a deep neural network, and obtaining training sequence data required by the deep neural network; performing a minimum mean square error channel estimation algorithm MMSE optimized by an iterative method to obtain frequency domain response CFR vector data at a channel pilot frequency, wherein the CFR vector data serve as input of the deep neural network; performing multi-layer deep neural network DNN to introduce a Sigmoid activation function is introduced, training an estimation network iteratively through training data, and acquiring the optimal output of each hidden layer neuron; calculating a difference value between a final actual output and a target value according to the output of each hidden layer, processing each difference value by using a cross entropy loss function, adjusting a weight and a threshold value of the neural network, and finally extracting an output symbol frequency domain response vector to complete channel estimation. According to the method, the CSI precision can be effectively improved, and the problems of time waste and precision caused by single training of a traditional deep learning algorithm are solved.

Description

technical field [0001] The invention relates to the technical field of wireless communication, and mainly relates to a deep neural network-based MIMO-OFDM system channel estimation method. Background technique [0002] With the continuous development of communication technology and the advent of the 5G era, both high-tech industries and civil and commercial applications have higher and higher requirements for communication speed and communication reliability and effectiveness. Multiple-Input Multiple-Output (MIMO) , MIMO) technology can effectively improve the utilization rate of space dimension and can effectively reduce energy consumption, improve spectrum utilization rate and increase communication rate, Orthogonal Frequency Division Multiplexing (Orthogonal Frequency Division Multiplexing, OFDM) technology can improve bandwidth utilization rate and effectively reduce Due to the influence of multipath fading, the combination of the two MIMO-OFDM systems can give full play...

AI Technical Summary

Problems solved by technology

But deep learning is now in the early research stage. Deep learning algorithms replace the interpolation process with deep learning. The data is often a single data, which cannot feedback the output of each neuron in a timely manner and there is a single training and The problem of spending a lot of time; most deep neural network algorithms do not consider the use of multiple communication scenarios, especially in MIMO-OFDM systems, due to the increase in the number of antennas, the estimated parameters increase and the parameter dimensions increase, thus making the neural network become It is complex and cannot effectively obtain training data for training iterations, which will increase the complexity of the algorithm and limit the performance

Images