To do this, you need to recognize speech, then parse recognized sentence, then launch user's command. you can use google text to speech engine. then, if text is "open file dialog" etc, open file dialog. also you can use windows speech recognition
on youtube[
^]. also the sentence paring thread:
Parsing sentence using template[
^]