A study on OFDM channel estimation based on inverse convolutional neural network

This topic takes Orthogonal Frequency Division Multiplexing (OFDM) as the research object to study the new idea of channel estimation using Inverse Convolutional Neural Network. This paper focuses on the in-depth study and improvement of the channel estimation method based on the guide frequency. Existing channel estimation algorithms based on the guide frequency mainly use the interpolation method to make full use of the channel characteristics in which the guide frequency is located, so as to obtain the complete channel state information. Compared to the interpolation method, a more accurate method is used to model the relationship between the position of the guide frequency and the entire channel. Using the channel simulation system of the University of Vienna, the collected signals are preprocessed and the deconvolutional neural network is learned offline using an offline method to reduce the channel fading. The results show that the algorithm proposed in this paper has better channel distortion and transmit signal detection compared to the MMSE algorithm.

I. Introduction

With the continuous development of communication technology, wireless communication systems need to face new challenges such as information transmission speed and large data processing, the application of orthogonal frequency division multiplexing technology provides a new type of technological innovation with high advantages in efficient transmission and high spectrum utilization. The complexity and characteristics of the wireless channel have an important impact on the accuracy of signal transmission, so the estimation of channel parameters is crucial. A reference signal approach is usually used to estimate the channel state, among which Least Squares (LS) and Minimum Mean Square Error (MMSE) are common estimation techniques [1]. With the extensive use of deep neural networks in wireless communications, they have significant theoretical and practical value in the fields of resource management, modulation identification, codecs, and signal detection [2].