T. V. Raman, a blind computer scientist who was trained as a mathematician, writes Thinking Of Mathematics —An Essay On Eyes-free Computing.
I think that even sighted mathematicians will get something from this, because the main issues for a visually impaired mathematician are that they cannot read or write in the usual way, and many of us do work in a situation where reading or writing is not available to us. Much of my best work gets done while walking to or from school, which is why I refuse to take SEPTA even though it would be faster. Plus, I get exercise that way. I've often taken to calling my own cell phone and dictating the solution to a problem into my voice mail. But this clearly isn't the same thing, because in the end I write things up in the traditional way.
Not surprisingly, Raman seems to find that the largest difficulties come in trying to communicate with other mathematicians, although this is becoming less of an issue as mathematics moves online, especially with the proliferation of TeX. (But this raises a question for me: often I write TeX that isn't strictly correct, but compiles anyway, and gives the right output on the page. How do systems like Raman's AS TE R (Audio System for TEchnical Readings, his Ph. D. thesis) handle this?