Having worked in the sound recognition field in mobile space (bird sounds) I know about this hard process a bit. Mostly I was using predefined algorithms and just implemented them in c#, but my poor knowledge and puny brain would never do such amazing stuff!
Natural UI comes to life! You can always see that some awesome MS Research powered speech recognition displayed on MS keynotes almost in real-time, but yesterday’s article on The Next Web highlighted the new addition to that wonderful feature: same voice tone speech translation in Mandarin! Check this out!