Research on peak and valley value prediction of RF envelope based on hybrid neural network
Aiming at the disadvantages of complex and computation-intensive control signal generation in traditional envelope tracking power supply,the paper proposes a hybrid neural network based on the series of fitting neural network and classification neural network to predict the peak and valley values of RF signal envelope,and simplifies the control signal generation method by extracting the key features of envelope tracking power supply.Firstly,a fitting neural network is used to predict the envelope of the RF signal based on the quadrature amplitude modulated mapping data.Secondly,three parallel classification neural networks are used to output the label data to predict the envelope of RF signal.Finally,the value of each data point in the predicted RF signal envelope is multiplied by its corresponding label data respectively to obtain the peak and valley value information of the RF signal envelope and obtain the required information of the control signal.The simulation results on Simulink platform show that the proposed method can reduce the computational load by 21.548%compared with the traditional method.Meanwhile,the classification accuracy of peak and valley value of envelope is 99.98%,and the root-mean-square error of predicted RF signal envelope is 0.183 04.
hybrid neural networkenvelope tracking power supplyfeature extractionpeak and valley values of the envelope
万方数据
维普
