Study On The Blind Equalization Algorithm Using Nerual Network

In modem communication system, the inter-symbol interference (ISI) caused by non-ideal character of channel is the main factor which effect communication quality. Blind equalization techniques rely on solely the received channel output signal to adjust the equalizer weights without a known training sequence available. In the fast mobile circumstance with high data rate, the absolute linear channel is nonexistent. So the research on the nonlinear channel blind equalization is now a hot spot in the current communication area.This paper is performed mainly on the blind equalization using neural network. The new ideas are proposed for structure of neural network, optimization of weights and cost function of neural network, based on summarizing the present study on the blind equalization using neural network domestically and abroad. The large simulations show that the algorithms put forward in this paper are effective.The main results of this paper are as follows:1. A modified genetic algorithm with maintaining diversity is presented for solving genetic premature convergence. And this algorithm is applied for optimizing weights of blind equalizer using neural network. The experiments show effectiveness of the algorithm.2. The blind equalizer of viable structure using neural network is proposed, we use elitism-based compact genetic algorithm optimize structure of neural network, this algorithm overcomes these shortcomings of complex operation, long searching time and large memory which appear when the simple genetic algorithm is used to optimizing structure of neural network. Then this structure-optimized neural network is applied to research on blind equalization, simulation results show this network for blind equalizer achieves the faster convergence speed and the smaller residual error in linear channeland nonlinear channel.3. This paper presents variable step dual model constant modulus algorithm, which combine with neural network for blind equalization in nonlinear channel. This algorithm improves convergence speed and reduces residual error by increasing small the quantity of computation.4. A novel neural Chebyshev orthogonal polynomial blind equalizer is proposed, which structure is simpler and convergence speed is faster than blind equalizer using neural network.