Mean Hand
40 points by hush
40 points by hush
This is cool. Some glyphs, like 'b' and 'h', are missing their ascenders, especially in the lighter weights. I wonder if it's a result of how the alignment or averaging processing was done. Also interesting that the dot of the 'i' merges with the stem, but 'j' manages to keep a separate dot.
Ironic that the average 'h' of the definitive OCR training set looks more like an 'n'!
Also interesting that the dot of the 'i' merges with the stem, but 'j' manages to keep a separate dot.
It is! I wonder if it's because simply omitting the dot is more common for 'i's than 'j's? I actually do the reverse is my own miserable handwriting (omit on j, use on i) on the post-hoc logic that lower and uppercase i, and lowercase l, need the dot more for disambiguation whereas placement of j and the upper case hat make it less prone to this. But super interesting observation in any case!
I noticed the same! Also, all of the letters with descenders are shifted up, so all letters have the same baseline. I imagine it is part of data collection.
Clicking through, this document has an example of the forms used to collect the data: https://s3.amazonaws.com/nist-srd/SD19/1stEditionUserGuide.pdf
The bold b and h have widening ascenders. It's definitely because people draw them in all kinds of directions. You could rotate the letter to align the ascenders?
What an amazing idea and the resulting typeface is surprisingly readable at regular weights. I could actually see myself using it.
Years ago I turned my own spider-scratch handwriting into a font and used it for a while on my website until the complaints got too loud to ignore. With this font you could at least say that you were using an average font.