High Power Amplifier Pre-Distorter Based on Neural-Fuzzy Systems for OFDM Signals

Single-Carrier Frequency Domain Equalization for Hammerstein Communication Systems Using Complex-Valued Neural Networks

“…If the nonlinearity is known and it is invertible, then the effects of can be compensated by inverting it. Specifically, an estimate of the transmitted data vector is given by (22) …”

B-spline neural network based single-carrier frequency domain equalisation for Hammerstein channels

“…During the data detection, given the estimated CIR vector h, the estimated nonlinear phase response Υ( ) and the estimated inverse nonlinear amplitude response A −1 ( ), the linear equalised TD signal w k can be computed according to (16), (19) and (20). The estimate of the transmitted data x k can then be given by …”

Digital Predistorter Design Using B-Spline Neural Network and Inverse of De Boor Algorithm

“…Existing predistortion techniques for compensating memory HPAs [5]- [17] can roughly be divided into three categories. The look-up table (LUT) based techniques [5]- [7] realize a PD by representing the inverse characteristic function of the memory HPA in a LUT.…”

“…By contrast, the direct-learning based PD designs [14]- [16] first identify the input-output relation of the memory HPA using a polynomial model and then adapt a polynomial PD directly to invert the resulting polynomial HPA model. A recent work [17] uses a neural-fuzzy based PD, instead of a polynomial based PD, in the indirect-learning structure. It is well understood that the memory HPA can be modeled by the Wiener model consisting of a linear filter followed by a memoryless nonlinearity [18].…”

“…The two types of static nonlinearity widely adopted to model the HPA are the traveling-wave tube (TWT) nonlinearity [18], [19] and the solid state power amplifier's nonlinearity [20]. Most of the existing researches dealing with the TWT nonlinearity, including [11]- [17], adopt a two-parameters output amplitude response model [19], which peaks at an input saturation amplitude. However, when the input amplitude increases beyond this saturation point, the output amplitude of this model actually starts to fall.…”