So I am trying to make an application to track potholes in a road. I am using accelerometer readings for that. Assuming that the phone is mounted in a car and our orientation doesn't change. After removing noise (through a low pass filter) I am saying that if I find a peak greater than a threshold in say the y axis I would like to classify that as a pothole. But I feel that this algorithm would confuse breaking with potholes, so how can I go about it.
Braking involves all the wheels while a pot hole involves one at a time. If your phone has a three axis accelerometer, like an iPhone, the acceleration vector will have a different direction profile relative to the phone for each phenomena. You should also expect a shorter time profile, ie higher frequencies for pot holes.
These should be distinguishable in the data you collect, so collect data and answer your own question