Find the previous alphabet.

khanji has a LOT of characters in their alphabet and are using a way of compressing it so that the most common characters are 1 byte and the less common are 2 bytes. If it's a '1 byte' character, it starts with a 0, If it's a 2 byte character, it starts with a 1. If you're at a particular character (at the beginning), how do you find where the previous character begins?

Now this is an interesting question.
Let the currect char be c (irrelevant), and the first bit of every previous byte be denoted by 1s, 2s....etc.

Some hints:
1. If 1s = 1, then previous char is 2 byte long.
2. If 1s = 2s = 0 then ??
3. If 1s = 0 then all (2...n)s = 1 then the problem is undecidable until you hit
a 0, in which case the answer is ??