Blind Equalization in Neural Networks

where m_ij(n) and σ_ij(n) are the expectation and variance of the space fuzzy domain, respectively, w_h(n) is the connection weight of the fifth layer.

5.3.2.2Fuzzy inference rules

CE(n) and E(n) are used as the controller inputs to control the step size of the FNN controller, then Gauss membership function is used to fuzzification inputs. The output is ∆μ(n). The iterative step size formula of equalizer tap coefficients is μ(n + 1) = μ(n) +∆μ(n),

The fuzzy inference rules are shown as follows:

(1)If E(n) is big and CE(n) is positive, then ∆μ(n) is PB;

(2)If E(n) is big and CE(n) is zero, then ∆μ(n) is ZE;

(3)If E(n) is big and CE(n) is negative, then ∆μ(n) is NB;

(4)If E(n) is medium and CE(n) is positive, then ∆μ(n) is PS;

(5)If E(n) is medium ...