I genuinely apologise for what may be a poorly worded question. I'm extremely tired but have a ridiculous huge and important project due in on Monday for my degree. Thank you in advance for any help or suggestions you may offer.
This is really frustrating me... I've had no problem teaching myself another programming language, basic filtering/resampling techniques and a few bits and bobs with transfer functions and transforms. Am I simply so tired (or dumb?) that I'm missing something obvious or is this a genuinely tough problem?
My problem is as follows:
I have a 3 axis accelerometer in my hand.
I have a 1 axis accelerometer that I'm standing on (... well a force meter/scales but you catch my drift as I know the mass of the system.) that "sees" in the vertical z axis.
What would be the best way to combine the data from the accelerometer in my hand so as I can compare against the one I am standing on?
I imagine a simple vector sum wouldn't cut it - acceleration-time plots for the hand accelerometer showed huge spikes due to x and y components when using this method that had almost no effect on the accelerometer I'm standing on.
I know pitch and roll can be calculated with an accelerometer... is this the pitch and roll around the axes if they were defined by the accelerometer itself or in another frame of reference? Would I even need the yaw?
EDIT: I think I remember reading something along the lines of the following?:
m_x = a_x + g*R(some euler angles?)
m_y = a_y + g*R(some euler angles?)
m_z = a_z + g*R(some euler angles?)
|m| = |a|+g|R|(some euler angles?)
Where m would be the measured acceleration from the accelerometer. a is the actual linear acceleration in a given axis... and R is a rotational matrix?
also:
pitch = arctan(-m_x/m_z)
roll = arctan(m_y/sqrt(m_x^2-m_z^2))
Considering I only want a_z... are there too many unknowns?