A Digital Predistortion Algorithm Based on Support Vector Machine of Mixed Kernel function

In this paper, a Digital Predistortion (DPD) algorithm based on the support vector machine (SVM) is presented. By introducing the kernel function, a complex nonlinear problem existing in the input space is able to transfer to a linear problem, which is mapped to a higher dimension space, so that a better predistortion is hopefully to achieve. Based on a new hybrid kernel function introduced in this paper, which is a combination of the Fourier kernel function and the polynomial kernel function, the learning and generalization ability of the algorithm used in the estimate of the predistortion function is improved. The simulation results verify that the performance of the method introduced in this paper is better than that of the traditional DPD algorithm based on Least Mean Square.