Communication Channel Equilibrium Theory And Application Of Technology

Wavelet neural network(WNN) has become a research hotspot as a novel approximal tool. It absorbs the merits of neural network and wavelet transform and trains network to adjust the shape of wavelets. It needs fewer base functions than wavelet decomposition and has faster speed of convergence than back-propagation(BP) network.In digital communication system, high speed of data transmitting results in inter-symbol interference(ISI). The receiver has to use an equalizer to overcome ISI. Traditional equalizer is the linear transversal equalizer(LTE) which has poor performance when the communication channel has nonlinear characteristics.In the presence of additive noise and nonlinear distortion, the main research work and results are as follows:(1) The linear transversal equalizer based least mean square algorithm(LMS-LTE) and the linear transversal equalizer based recursive least square algorithm(RLS-LTE) have been studied. LTE has the follow shortcomings: slow convergence-speed, high-rate bit error.(2) A new initial method has been adopted to reduce the size of the network. The new wavelet neural network is applied to channel equalization and has been compared with LMS-LTE, RLS-LTE, and the equalizer based wavelet neural network(WNNE) in nonlinear channel.(3) In order to increase the speed of convergence, we use entropy function as penalty function. And the network adopted the algorithm with momentum gene and varying learning rate to correct weights of network and minimize error of approximation.