The Research Of Blind Equalization Algorithm Based On The Fuzzy Neural Network

In the wireless and digital communication systems, the Inter-symbol Interference (ISI) is often encountered because sending signals are influenced by complicated transfer mechanism. In order to reduce ISI, the equalizer has to be used in the receiver to compensate the channel's characteristic and renew transmitting signals accurately. The traditional equalizers need to send the training sequence constantly, so it can't satisfy the developmental request of the digital communication technology. Blind equalization technology only utilizes the statistical property of received signals to adjust parameters adaptively without the assistance of the training sequence. That is, blind equalization technology can start-up to convergence itself and prevent the lost-lock situation without any training sequence, and it can make the output of the filter be approximate to the input of the sender. Blind equalizer overcomes the disadvantage of the tradition equalizer, it can trace the channel characteristic and estimate signals successfully in despite of the channel influence such as frequent fading, serious unlinearity, time variety, multi-path spread and so on, even the communications halt because the receiver can't trace the channel characteristic. Blind equalization technology has been applied to many areas such as communication, radar, earthquake prospecting, image processing and so on.The major contribution of this paper is summarized as follows:1. In this paper, the principle of blind equalization algorithm and the development status quo of blind equalization technology are analyzed systematically. The peculiarities, structures and establishment methods of the fuzzy neural network (FNN) and selecting rules of membership functions are researched. And learning algorithms of FNN are expounded in this paper. Three blind equalization algorithms based on the FNN are proposed and their realization principles are analyzed in this paper.2. The design principles of variable step-size blind equalization algorithm is introduced after analyzing fixed step-size blind equalization algorithm, because the step-size can affect the convergence result. In this paper, blind equalization algorithm based on the FNN controlling step-size is proposed, and it using the control functions to improve the control precision of step-size. This algorithm controls the value of step-size real time using error, error change and correlative rules. It adopts big step-size to improve the algorithm convergence rate in the beginning, and adopts small step-size to improve the algorithm convergence precision when the algorithm is converged. It can solve the contradiction between the convergence rate and convergence precision.3. The design principles of blind equalization algorithm based on FNN controllingαfactor is proposed after analyzing the affection ofαfactor in forward neural network blind equalization algorithm, because nonlinear modifyαfactor affects the convergence performance. In this paper, the error and error change as inputs of the FNN, the control algorithm is used to get the variableαfactor. It adopts bigαfactor to quicken the algorithm convergence rate in the beginning, andαfactor monishes gradually along with the degree of convergence to recede the spare error and optimize algorithm.4. The principle, classify, trait and application of fuzzy clustering are analyzed in this paper. Then a new blind equalization algorithm based on the FNN classifier is proposed according with the characteristic of FNN clustering. The traditional blind equalization technology uses decision to resume sending signals. It adopts threshold and belongs to the hard decision. In this paper, The FNN classifier replaces the decision to soft decide and improves the right rate of signals.5. The principle and classify of the dynastic fuzzy neural network (DFNN) are researched in this paper. In the wireless communication system, the process of blind equalization is a dynastic process because of the time variety and uncertain characteristic of the channel. In the equalization process, DFNN accords with the characteristic of dynastic channel because it can use the current and history data. A new blind equalization algorithm based on the DFNN replacing landscape orientation filter is proposed.