The data doesn't seem representative of UTF-16, UTF-8, UTF-7, or ASCII when interpreted as binary either by assuming the dots to be 1s or 0s. I've experimented some more with 7-bit bytes and still nothing. I see the probability of a binary character encoding as even less now than I did initially. The dots may encode characters, but perhaps using a different method.
Here is one method that could have been perfect if it actually worked (lol):
Night Writing. This was a predecessor of braille that, according to the linked Wikipedia article:
It was designed by Charles Barbier in response to Napoleon's demand for a code that soldiers could use to communicate silently and without light at night.
What better encoding for an ODST soldier to encounter at night? The encoding requires two columns six-dots high in order to represent a single character. Among the three J-banners we have 12 rows of 28 dots. This could result in two rows of 14 characters in "Night Writing". If only our data fit...
The thing about night writing is that the columns of dots are always contiguous; that is, there is never a need for there to be blanks between dots when scanning from top to bottom. If you look at the three blocks of binary we have, you can see that there is no possible way to order them top-to-bottom such that you end up with two valid 6-dot "Night Writing" columns on top of one another in the resultant leftmost column of 12 dots.
So I have dropped that theory.
Then I looked at
Polybius Squares. This has merit, but what is our square and how is it encoded in the dots? My initial mappings have failed. Why would three dots in a column be expressed in multiple ways if the pattern wasn't significant?
This brought me back to the notion that the 4-dot columns encode either a number from 0-15 (or 1-16) or two numbers from 0-3 (or 1-4)... in binary. So I'm back, full-circle, to binary... but then what?
I really like Moose's idea of overlaying the data from the three banners, perhaps to construct a "fifth row". We may also need to consider numbers we have seen elsewhere, such as the
Optican Image. Perhaps 49.2.7 plays in here somehow. I had even considered that the barrier codes helped define a Polybius Square. As always, I'm left with more questions than answers.
Back to the drawing board...
