If you have seen it work with Tesseract, then it should be possible with IEVision. Are you able to find any 14 segment Tesseract training sets?
Of course, you should also consider the other possibility of passing each character image to a method that checks on/off state of each of the 14 segments to determine the character.