I am doing a graduation research and got stuck for two weeks solving this problem. I am not a expert in mathematics so I might use the wrong expressions.
I have sensor data that measures the movement of the x, y, and z axis. Basically, I am only interested in the amount of movement/activity. So I do not need to know any actual rotation or translation/movement.
And I want to divide this movement into 2 groups: movement and fixed. I have observed if the sensor was moving or not. Here is a example of my data:
And as you can see it is obvious when the sensor was mostly moving and when it was mostly fixed. Now I am trying to make values that significantly distinguish both groups. And this calculated values should fit in a normal distribution that do not overlap. This would be a acceptable result, although it would be better when they do not overlap at all:
And I have tried many things to get the wanted result but none of them work. Here are some examples of what I tried.
- absolute gradient x axis + absolute gradient y axis + absolute gradient z axis.
- absolute rollingsum x + absolute rollingsum y + absolute rollingsum z
absolute gradient x axis .....
But no matter what I try, I do not get distributions that do not overlap (a lot).
Does anyone know how I can get values that will result in not or just a little overlapping distributions?
I hope I was clear. Greetings