Genetically optimised feedforward neural networks for speaker identification