It comes down to defining "simple". Agreed measuring back-emf is more complicated than simply measuring applied motor voltage. But like many things in life you get what you pay for. And of course the objective is to detect deceleration...not speed. I suggest waiting until the OP gets closer to implementation when we can hear exactly the configuration, command vs. conventional, whether soldering and component-level assembly is on/off the table, Arduino on/off the table, and so on.
As for the hanging pendulum inertial sensor, this would be like those cattle cars that would trigger the cow or chicken sounds upon movement... IIRC, QSI came up with this sometime in the last century.
Clearly this would be a simple and inexpensive method to detect deceleration. The trick would be to only detect deceleration and not acceleration. In a cattle car trigger one expects sounds upon starting or stopping. Likewise, if used for curve detection it doesn't matter if "turning" left or right. So it would take some thought to "tune" the pendulum mechanism or insert an electrical filter circuit to only generate a trigger when it swings in one direction so to speak. That's the advantage of chip accelerometers as the polarity of the voltage output indicates both the sign and magnitude so that acceleration can be ignored.
