@RayDunzl -- Yes, I think those are conventional microspeakers. Some use piezoelectric drivers. Not MEMs.
@Cosmik -- One of the fundamental issues. Like planar designs (ESLs, planar dynamics), displacement is limited, so you need a lot of them (large area) to achieve high SPLs. A lot means high yield, high cost, etc. And it takes a bit to drive them all since you are generating enough field strength to deflect a mechanical beam (beam like on a teeter-totter, not like an EM or optical beam) on a whole lot of them at once. They can also be delicate and with finite lifetime due to essentially mechanical wear. I think, not claiming to know all that much about MEMs.
One scheme I vaguely remember used a carrier approach, many MEMs (or other drivers) oscillating with the array modulated to produce interference patterns that produced sound. The implications for programmable beam forming, dispersion characteristics, etc. are interesting. One claim was that they could "place" the sound at a specific place in space and away from that you might hear nothing or noise. Never heard if it went any further and a job change took me away from those sort of research articles so I've not kept up.