A Compensation Method Based on NN at the Transmitter and the Receiver Level for Nonlinear Distortion in MIMO OFDM Sytems Using MMSE Receiver

We propose a Non-Linear (NL) distortion compensator based on Neural Network (NN) accompanied with MMSE (Minimum Mean Square Error) which corrects the nonlinearity introduced by HP A (High Power Amplifier). The neural network consists on a feed-forward Multi-Layer Perceptron (MLP) associated with Levenberg-Marquardt learning algorithm. In this paper, the correction is done at the transmitter after HPA and at the receiver level after OFDM demodulation. We present the structure of the NN adapted to both the predistortion (NN-PD) and frequency domain. Finally some simulation results are presented in a VBLAST MIMO OFDM (Vertical Bell Laboratories Layered Space-Time Multiple-Input Multiple-Output Orthogonal Frequency Division Multiplexing) system running under a Rayleigh fading channel.