You can use Encode::Guess, I believe, which performs a guess. To be a good guess, you have to prime it with the most likely charsets (e.g. if you are expecting English, probably ASCII, ISO-8859-1, etc., while if you are expecting Japanese...).

However, here is a shameless plug for a project I'm tangentially working with (though not leading): HEBCI, or HTML Entity Based Codepage Inference, though I often forget that and think it's Heuristic Estimation of Bytes' Charset, Idiomatically, or such.

HEBCI is the way to figure out a charset by sending some stuff that comes back differently in different charsets, and checking the differences. HEBCI is an HTML way to do this, but I imagine that you could try the same principle with email, in the event that Encode::Guess doesn't suffice.