| Hi-index | 0.00 |
In the conventional method to design recurrent RBF networks for channel equalizers, firstly, the impulse response of the channel is estimated with an adaptive FIR filter. Secondly, all noise-free received signals are estimated with an estimate of impulse response. However, the performance of the network designed with this method is degraded down if the channel is nonlinear. In order to overcome this drawback, we apply Miyake's (Miyake et al. 1996) method to the conventional training method of recurrent RBF networks. In the proposed method, we estimate the received noise-free signals with the relation between the training signals and the received signal. Then we design the recurrent RBF networks with the noise-free received signals.