| Summary: | 碩士 === 國立雲林科技大學 === 電子與資訊工程研究所碩士班 === 92 === In order to reduce the inter-symbol interference caused of the signal transmission through wireless channel, an efficient equalizer used to recover the distorted signal is highly required in the receiver. Conventionally, linear equalizers partition the signal space into some decision region in a linear manner. However, the problem of the optimal equalization is indeed a nonlinear issue. In the literature, Bayesian theorem, radial basis function network, and fuzzy filters are proposed to achieve a nonlinear equalization.
In the thesis, we propose a new equalizer that is based on a combination of neural network and fuzzy inference system. The advantage of the fuzzy inference system is that does not need the precise mathematical model, and may unify humanity''s knowledge in the system design. It is possible to enhance the learning capability of fuzzy inference system by combining the neural network with the back-propagation algorithm (BP). That enables the proposed Neuro-Fuzzy equalizer can update its system parameters in accordance with the time-varying behavior of the wireless channel, and hence, the proposed equalizer achieves better system performance and lower bit error rate.
|
|---|
