The Internet is full of lexical analyzers, as this was one of the favorable hobbies before the arrival of html and windows widgets: you are re-inventing the wheel. How about lex(1), flex(1), or even a full parser like RecDescent (if you don't mind the slow speed.)

Lots of analyzers out there. Many parsers can also act as lexers, and some (rich) lexers like flex(1) can almost act like parsers. There is also a perl binding to flex(1) through my cpan module Parse::Flex.