Speech recognition: Past, present, and future