Study On Blind Equalization Algorithm Based On Wavelet Neural Network Theory

In modern communication system, the inter-symbol interference (1SI) caused by non-ideal character of channel is the main factor which affect communication quality. The conventional equalizer resort to a training sequence in order to overcome inter-symbol interference. But this method would influence communication efficiency. So it is currently a hot research issue in the field of digital communication. Blind equalization techniques rely on solely the received channel output signal to adjust the equalizer weights without a known training sequence available.The wavelet neural network is an improved neural network structure by making full use of the advantages of wavelet transform time-frequency localization and neural network self-learning property, which has the improving faulty-tolerance and better property in accuracy. The wavelet neural network has been well applied in many areas, for example, signal processing, data compressing , system identification, and so on.The major contribution of this paper is summarized as follow:1.This paper systematically describes the basic theory of blindequalization based on neural network, and sums up the advantages and disadvantages of the conventional algorithm. Then we propose a new blind equalization algorithm based on wavelet neural network by making full use of the advantages of wavelet transform time-frequency localization and neural network self-learning property.2. The first algorithm based on feed-forward wavelet neural network, this algorithm is developed by the combination of conventional CMA and feed-forward wavelet neural network. The second is blind equalization algorithm based on bilinear recurrent neural network. This algorithm adopt the bilinear recurrent neural network and design a new improved neural network structure. The paper also gives the iteration formula of the algorithm.All proposed algorithm have excellent performance not only in real-valued system but also in complex-valued system.3. The computer simulations show that all proposed algorithm have more fast convergence performance than traditional algorithm.