In order to record audio on a web client, and send that back to a cgi script on a web server, you'll need javascript, I think, to actually control the client's sound card and to use client-local disk space or memory to hold the audio data long enough for transmission back to the server.

As for what you do with the audio data once it's back on the server, you'll need some speech recognition software, which is not written in perl (though there are a couple of CPAN modules that provide wrappers for working with speech recognition software -- assuming you can get such software to run on a web server, and don't count on that).

In other words, this is not a perl question, per se.

(Trust me, just using buttons on a form will work a lot better.)