Best technique for detecting charset/character encoding of RSS feeds

I have a Ruby application that fetches RSS and Atom feeds. I’ve tried
using the chardet gem (UniversalDetector) to figure out what the
character encoding of each feed is. But for some strange this library
thinks a lot of feeds are EUC-KR (Korean) when they plainly aren’t.

Can anyone suggest a better way to find the encoding of RSS and Atom
feeds (e.g. via BOM detection, etc.) with Ruby?