IMU's to estimate distance over short periods. Can drift be ignored?

Hi, I want to measure time taken to travel a distance of 5 meters. Ideally I would like a single handheld device instead of having devices on the wall shooting ultrasounds or whatnot (that might be what I end up doing however if the IMU option is not viable).

So I was thinking estimating a 5m distance with an IMU might not be too bad when thinking about drift. My main concern would be: Is it going to be repeatable? If I use the device to estimate time taken at various speeds, will the IMU estimate a relatively constant distance (even if it drifted from the real 5m) let's say, would it always be around 4,8m, or else are IMUs very noisy and unpredictable so it would estimate 4,2m on one try, then 5,9m on the other?

If you guys think an IMU will drift too much even after only 5 m, would you guys have another idea of how walking speed could be calculated? let's say it is indoor and should be battery powered. Thanks!

You can read about the problems with that idea here: http://www.chrobotics.com/library/accel-position-velocity

IMUs drift over time, distance is largely irrelevant.

Thus through integrating acceleration over time, one has a far better idea of when a thrown ball has traveled five meters than when a tortoise has traveled the same distance.

So IMUs are clearly not suited for measuring distances over 10 seconds. I need to find some other technology. I was thinking maybe 2 sensors on a wall in the hallway that shoot at people and then we calculate their speed with trigonometry. But I need to target specific people and ignore the others passing by. Maybe the targets could carry an object that communicates with the sensors? what are your thoughts?

I need to target specific people

That is the hard problem, so solve that first. RFID tags or cameras and computers with face recognition come to mind.