I was too thinking that based on the title, but it looks like that the goal is a little bit different and indeed quite interesting:

The idea here is to harvest the energy from the "RF" ambiant noise, not from a pre-provided relatively powerful signal like the one that is powering RFID tags.

Based on what says the article, the power level of the "ambient" energy is too low to be harversted by current technology except the one presented in this article.