Channel blind equalization method and system based on Bagging and long and short-term memory network

Abstract

The invention discloses a channel blind equalization method and system based on Bagging and a long and short-term memory network, and the method comprises the following steps: building a channel blind equalization model based on a long and short-term memory network; adopting a Bagging algorithm to optimize the blind equalization model of the long and short-term memory network; training the obtained model to obtain optimal parameter setting; and balancing new data by using the trained blind equalization model. On the premise of ensuring good fitting of the neural network to nonlinearity, the problems of difficulty in parameter optimization and easiness in falling into a local optimal solution in the blind equalization process of the neural network are solved by introducing the long and short-term memory network. Besides, the influence of a channel on the amplitude and phase of a transmission signal is eliminated based on an improved norm algorithm MCMA, so that the technical scheme of the invention is more adaptive, the applicable channel environment is wider, and the equalization performance of a neural network type equalizer in a time-varying channel can be improved.

Description

technical field [0001] The invention belongs to the technical field of channel equalization, in particular to a blind channel equalization method and system based on Bagging and long-term and short-term memory networks. Background technique [0002] In the wireless communication process, each symbol is transmitted according to a certain time interval. However, in this process, the received symbols often overlap due to factors such as unsatisfactory frequency response of the wireless transmission channel, multipath effects, and noise. This phenomenon is called inter-symbol interference (inter-symbol interference, ISI), which is the main factor affecting the correctness of the received signal. In order to reduce the impact of factors such as inter-symbol interference on the received signal and improve communication quality, an equalizer is usually added at the receiving end to make up for it. This process is called channel equalization. [0003] Channel equalization technolo...

AI Technical Summary

Problems solved by technology

[0005] The present invention aims at the problem that in the prior art, when channel blind equalization is introduced into a neural network instead of a linear filter to construct an equalizer, the convergence speed is slow or it is easy to fall into a local optimal solution; Long short-term memory network, and provide additional information for the current equilibrium to further improve the equilibrium effect

In addition, the MCMA algorithm is used to construct the loss function to solve the influence of the channel on the amplitude and phase of the transmitted signal

Images