All the positive acceleration happens in a very short time. It might be hard to measure it frequently and accurately enough to use the integrated results as a velocity estimate.

Perhaps measuring the negative acceleration caused by air friction could give you a velocity estimate (the velocity-to-friction function is probably non-linear). But this may vary a bit with air density, temperature, etc.

I'd be interested to see how it's done. My bet is that it's not directly measuring a near-impulse acceleration.