Or tune your radio (or whatever) to an unused frequency and let the white noise rule.
Or turn on a room fan to rustle flexible stuff around.
Actually, what's scary is the processing they came up with. Given that, you wouldn't need a laser bounced off a windowpane. Just focus on a reflection of a stationary outside object.
Processing isn't all that complex
For high speed camera (frame rate >> sound frequency) its simply region summation, difference from temporal mean, than FFT over a bunch of frames, with a sliding window, then weighted frequency filter, then IFFT.
For low speed, its region selection (the area with the bag), summation of each row, difference of each row to region average over a bunch of frames, then FFT, then weighted frequency filter, then IFFT.
Its a few lines of matlab. The hard part is the target. The reason why a bag works is its a neat combination of high reflectivity/complex shape (ie signal changes substantially with even small motion) and really thin (even very low pressure sound makes it move at usable frequencies).
For thicker stuff (window, etc) you need either an interferometric measurement or the surface is both planar enough and specular enough to cause a small but measurable (as the receive spot moves) angle change even at very low deformations. Close range laser mics use the latter, long range use the former.
Interesting thought experiment. With a laser dot that is 1mm diameter, at 10m, a 1% difference in detected signal corresponds to a slope of 1:500,000 roughly, meaning even if the window was a yard across, if the center moved less than a micron (<1:50th of the width of a hair) at the middle relative to the edge, it would be detectable.
For long range, its harder because the air motion and beam spread makes the angle/intensity correlation almost impossible. But there, since the air effects are larger (10's of microns of path change over 100's of meters) and the frequencies lower than most sound (100's of Hz, and white-noise dominated), a "fringe velocity" type measurement with good post processing can make interferometric methods viable (but requires a really good laser source, ideally multiple spots to eliminate speckle, and a few other things)...and they do work, but AFAIK you can't do it off something with limited deformation (thick metal wall), or with internal damping (brick, drywall, etc) or with limited reflection. So glass tends to work the best (reflective, thin enough to have deformation, stiff, limited damping).
It doesn't take much sound...60dB is 0.02 Pa, more than enough to get microns of displacement in a decent window even at audible frequencies
