Power amplifier digital pre-distortion method of complex-valued full-connection recurrent neural network model

Abstract

The invention discloses a power amplifier digital pre-distortion method of a complex-valued full-connection recurrent neural network model. According to the method, a complex power amplifier model issimulated through a complex-valued full-connection recurrent neural network model, a power amplifier inverse model and achieving the adaptive digital pre-distortion is realized. The power amplifier model is based on a complex-valued neural network theory, an improved complex-valued real-time recursive learning algorithm is adopted based on a real-time recursive learning algorithm; and more accurate model approximation is realized for the power amplifier model of a digital communication system transmitting terminal. According to the method, the real-time recursive learning algorithm in the recurrent neural network is combined, a complex-value full-connection recurrent neural network model with a better effect is provided based on an original real-value recurrent neural network model, so that the complex-value real-time recurrent learning algorithm is further popularized. Through simulation verification, the model structure and algorithm are good in performance in the aspects of training time and modeling accuracy, and the high fitting degree of nonlinearity of the power amplifier can be guaranteed.

Description

technical field [0001] The invention belongs to the technical field of digital signal processing of wireless communication systems, and in particular relates to a digital predistortion method based on complex-valued neural network power amplifier modeling. Background technique [0002] In modern communication systems, due to the dual pressure of high-speed data transmission requirements and limited spectrum resources, in order to improve spectrum utilization, modulation methods such as Quadrature Amplitude Modulation (Quadrature Amplitude Modulation, QAM), Quadrature Phase Keying (Quadrature Phase Shift Keying, QPSK), Orthogonal Frequency Division Multiplexing (Orthogonal Frequency Division Multiplexing, OFDM), etc. are gradually widely used in communication systems. However, this type of modulation technology will increase the design difficulty of RF power amplifiers. This type of signal is an envelope modulation signal with a high peak-to-average power radio (PAPR), which ...

AI Technical Summary

Problems solved by technology

[0004] The technical problem solved by the present invention is: in view of the problems of insufficient model prediction accuracy, high computational complexity, and insufficient pre-distortion correction ability of the existing neural network-based power amplifier modeling method, the traditional real-valued neural network model and training Based on the algorithm, a power amplifier digital pre-distortion method based on a complex-valued fully connected recursive neural network model is proposed, which can establish a power amplifier behavior model more accurately and fit the nonlinear curve of the power amplifier well, that is, it has a very good model accuracy. degree, on the basis of higher model accuracy, the method of inverting the power amplifier model can be used to obtain a predistorter with good predictive effect

Images