What my personal top changes list would include are:
1. Much higher frequencies than 1-4G. We are talking in the 10s of GHz for 5G. This allows for a very significant increase in channel bandwidth possibilities, compared with frequencies in the 800 MHz to 1900 MHz region.
2. Filter Bank Multi Carrier (FBMC) modulation, most likely, as opposed to OFDM or spread spectrum. This increases spectral efficiency and allows for multiple simultaneous clients to share the channel, by assigning subcarriers to different clients.
3. Massive MIMO, which is really a different paradigm than the MIMO we have seen so far. Client devices may only have one receive antenna, for instance, but the base station, with a huge number of transmit antennas, can localize individual clients in space, using beam forming.
As Albert already noted, you can't expect 5G speeds at lower microwave bands. In that sense you also need a marketing layer between a physical layer and user layer, that will use psychological power to convince a user that 10Mbps is 5G /sarc.
IMHO mm wave technologies are a way to go. 60GHz pops to mind, as there is a very beneficial oxygen absorption which limits cell coverage in a sense that distant devices can't interfere, so network planning is very relaxed. More room above, but not with perfect absorption.
I also believe the modulation scheme must be simplified to facilitate complex MIMO. It is not that a complex modulation scheme can not perform in the air, it most certainly can, but to put it there you need to crunch it all in real time by some speedy processors, and those guzzle lots of energy. Also the more complex scheme, higher the peak to average ratio at each TRX, hence even more energy is wasted.
We are talking about low energy devices, most probably powered by some crazy environmentally friendly source, which means there is not much energy at hand. so they must stay low energy throughout.
Operating frequency-wise, there seem to be two camps, those operating below 6GHz and those above. For large cells we will need to operate below 6GHz to get the range. In femtocells (i.e. a single room) people are talking about 28GHz or around there. At the higher frequencies there is greater bandwidth to play with, but also greater pathloss. Beam forming may be a solution, or massive MIMO. No one really knows yet which is why pretty much anything new and cool can be described as 5G.
Furthermore, the standardization bodies 3GPP with the Release 12 (and behind) is ongoing in the Physical Layer specification for adaquate 5G radio Technologies
Behind this, the ITU called their 5G related standardization activities as "IMT 2020" and you can found some relating information in the following links