Nonlinear Channel Equaliser using Discrete Gabor Transform

Abstract

The adaptive equaliser makes use of adaptive digital filters whose filter coefficients are modified depending on the channel characteristics at the front end of the receiver. The noise introduced in the channel gets nullified and hence the signal-to-noise ratio of the receiver improves. Discrete Gabor Transform (DGT) helps decorrelate input data because of which the convergence speed of the Mean Square Error (MSE) improves considerably.  It is found that the time domain LMS equaliser is slow in convergence. To improve the convergence rate and MSE floor level transform domain-using DFT, DGT and DWT (2, 5, 6 and 7) has been studied. It is found that all the orthogonal transforms perform similar in convergence rate and MSE level. This paper aims at the evaluation of channel performance using Gabor transform, which is both frequency and time domain transform. The nonlinear channel equaliser using Discrete Gabor Transform is reported in this paper. Though Gabor Transform is a nonlinear transform as well as non orthogonal transform it is expected to be better fitted for nonlinear channel. Gabor transform based adaptive equaliser, though has a longer training time, has been found to have better noise recovery property and Lower MSE level, especially when the additive noise in the channel is large.