The thing is built with most of its innards able to move slightly within its case. Between aforementioned innards and the case itself are piezocrystals which, when compressed, put out a small electrical charge. It's not a lot on any single compression, but if you keep the thing in your pocket(book) or in a holster, so that it moves when you do, it's intended to be enough to keep the battery charged.
What this reminds me of is the original self-winding wristwatches. Before the dawn of recorded history, people had to actually remember to wind their watches on a regular basis (imagine!). Then came the self-winding watch which, if you wore it on your wrist and were normally active, used the impetus from your arm motions to incrementally recompress its own mainspring. The Nokia technology is entirely different, but the principal is the same -- capture energy which is being expended anyway in the normal course of events, and apply it to powering an existing application.
That's an important principal in terms of optimizing energy efficiency. Many devices waste a lot of energy (often exhausted as heat, but not always). If we can think outside our traditional parsimonious design models, we can surely find many more ways to utilize energy (often in forms we're not used to thinking about) that's currently going to waste. Cogeneration is a classic example, perhaps even micro-cogeneration powered by exhaust heat from industrial equipment or motor vehicles.
Staying within the consumer electronics sector, Nokia has set an example that lots of firms should be able to follow. Consumer electronics are responsible for a large portion of the increase in electrical use in North America over the past couple of decades. For stationary devices, a good portion of this can be eliminated by use of smart surge suppressors/power strips. But more and more electronic devices are becoming "personal" in the sense of mobile. And for that set of applications, something akin to the Nokia approach seems very attractive.