Speaker independent recognition of continuous speech by means of neural networks and dynamic programming.