Blind Equalization by Kernel Method with Variable Step Size for Underwater Acoustic Channel

To solve the problem of blind equalization for nonlinear underwater acoustic channel, the kernel method was adopted and a variable step size algorithm was proposed to further improve the performance. The kernel least mean square (KLMS) algorithm can obtain good performance for nonlinear filter with simple computational complexity. The step size is an important parameter of KLMS. Hereby a variable step size algorithm based on the estimation of instantaneous gradient error by sliding window method was proposed. The step size keep big value when the estimation value larger than the pre-set threshold, otherwise the step size switch to small value. The proposed algorithm can take full advantage of big value to obtain fast convergence rate and small value to obtain high convergence precision. The simulation results under nonlinear underwater acoustic channel show the effectiveness of the variable step size KLMS blind equalization.