Regarding the footsteps technique.
you can choose every sound in "Group" but the only sound what you can hear at the moment is "Dirt no matter what sound you choose.
Not true. You can choose any sound and it will be audible. The problem is to make different regions sound different. To make it work you need to add another action to the responder state, triggering the other detector (I consider the case of two detectors). So you have RegionSensor1 and RegionSensor2, and corresponding responders Responder1 and Responder2. Add EnableDetector action to the current (1 by default) state of each responder, in parameters select the opposite detector, ie RegionSensor2 for Responder1 and RegionSensor1 for Responder2, and that's it. Ah, and another thing - check the "Trigger on exiting volume" box for each of the detectors. You may also want to make the detector geometry high enough so jumping won't trigger it.