Speech recognition: a hybrid approach