Hello there, Now I will give you what we call in sweden a "tankenöt",( brain teaser?)
Do you think it's possible with some kind of algorithm to calculate the distance or the direction to the source of a footstep/knock from an accelerometer? I mean if you can get data from x, y and z axis from the accelerometer wouldn't it be possible?
I was thinking of the ADXL 345 accelerometer to be more precise 
thank youu!
You may be able to tell about where footfalls are landing based on the magnitude and direction of the vibrations they create. The hard part will be gathering data and recognizing patterns.
Very difficult to map an event that might only be 100 - 300 ms long and relatively low in amplitude. Almost easier to try and map it with three audio sensors from the time delay and amplitude in the arrival to the sensors. The real issue is the speed of propagation through different materials to a single sensor and that a single point of reference might not be sufficient to return good data as although you do have some sensitivity, the event is low in energy/amplitude and to try to amplify it would open up a really big can of worms. I don't think that the sensor in question was designed for that purpose.
Doc
i think this is exactly what is already done with earthquake; you need at least 2 sensor to know where the things happened in a plane (so 2d) and 3 to know info int the space (so 3d).
with earthquake the "roundness" of the heart is enought to have a significant height difference. Also it is almost impossible understand thinghs when too much is going on; backreverese the "sum of effect" to the original effects need many more sensor and complicated math
From the description i'm not quite sure on the application but be aware that:
Displacement is the double integral of a Raw Acceleration in a given time period.
Measuring displacements with accelerometers is possible yes.
But, please consider the behaviors differences of piezo vs strain gauge accelerometers in low to high frequencies. (Ie, piezo types are not responsive to low frequencies (>50Hz) or changes in direction, if you are considering measuring a humans foot step it can be assumed that one footstep =~1Hz, this is low and I very much doubt peizo accels would give you any good data.)
Kind regards,
Do you think it's possible with some kind of algorithm to calculate the distance or the direction to the source of a footstep/knock from an accelerometer?
Here's a thought experiment for you. If you know the volume of a brick, do you think it's possible come up with an algorithm to calculate the length, width and height of the brick?
It''s not a brain teaser. It's a "Do you have some particular knowledge" question?
Do you know and understand vector math differentials and integrals? Is this just about decoding dynamic position as opposed to static position? The the answer is yes and you can do it.
So yes -- there are seismic algorithms as has been pointed out. They are an analog to what you want.
Google this, First Person Shooter with a real gun. Quote:-
The goal of this project is to be able to triangulate the position of an impact on a wall and then generatea mouse click that corresponds to this location. A computer screen can then be projected onto the wall and a projected action can be performed - seeming as if the impact acts as a mouse click on theprojected surface. This can then be extrapolated into use in any software, in our case a flash version ...
is the principle of earthquake control station (i don't know their name)