Abolfazl Falahati - Department of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran
Sajjad Nasirpour - Department of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran

AbstractThe disparity introduced between In-phase and Quadrature components in a digital communication system receiverknown as I/Q imbalance is a prime objective within the employment of direct conversion architectures. It reduces theperformance of channel estimation and causes to receive the data symbol with errors. This imbalance phenomenon, at itslowest still can result very serious signal distortions at the reception of an OFDM multi-carrier system. In this manuscript,an algorithm based on neural network scenario, is proposed that deploys both Long Training Symbols (LTS) as well asdata symbols, to jointly estimate the channel and to compensate parameters that are damaged by I/Q imbalanced receiver.In this algorithm, we have a tradeoff between these parameters. I.e. when the minimum CG mean value is required, theminimum CG mean value could be chosen without others noticing it, but in usual case we have to take into account otherparameters too, the limited values for the aimed parameters must be known. It uses the first iterations to train the systemto reach the suitable value of GC without error floor. In this present article, it is assumed that the correlation betweensubcarriers is low and a few numbers of training and data symbols are used. The simulation results show that theproposed algorithm can compensate the high I/Q imbalance values and estimate channel frequency response moreaccurately compared with to date existing methods

I/Q Imbalance; OFDM; Zero-IF; Direct Conversion; Neural Network; Channel Estimation, FrequencySelective Channel