Tri-axial accelerometers

[Irini]

Hello all,

 

I have a question and I hope some of you might have an answer for me.

I have to combine the 3-axes signals from a MEMS accelerometer in MATLAB so that I have the resulting signal as an input to several filters.

Do you know how is that possible? I tried to find more information online but that was not successful.

Whatever you know about this topic is useful so please do not hesitate to contact me.

 

Eirini

[D H]

What are you trying to use the accelerometers for? Your question is a bit too open-ended; you are asking us to write a book (multiple books!) and teach you the contents of an upper level undergraduate / graduate level class (multiple classes!).

 

Do you know what a Kalman filter is?

[ewmon]

While I have worked with three-axes accelerometers, I have not done so in MATLAB, but D H does broach valid concerns. What is the purpose of your work? What advantage would be the sensed acceleration in spherical or cylindrical coordinates compared to Cartesian coordinates?

[Irini]

Hello again,

 

I do know about the Kalman filter although it is not the one that I am going to use.

 

I am getting the motion artefact from hand/arm vibration in the three dimensions (x,y,z). The filter that I will apply is the LMS (least mean square algorithm).

One way is to use a 1-axis accelerometer to capture the motion (but I know how I can apply this since it is only one input) and the other way is to use all three axes (x,y,z) signals.

Now, I can have 3 separated outputs - one for each dimension - or to combine the three input accelerometer signals to have only one input for the LMS adaptive filter.

My problem is how can I combine those three signals to have only 1 input that will represent successfully the acceleration in all three directions.

 

Thank you all for responding

[ewmon]

It seems it's not a matter of translating the sensed motion from one Cartesian system to another.

 

Combining three signals into one input? There's the magnitude of the resultant, but that alone fails to fully quantify the motion, unless you can translate the output of the filter into the three axes by using θ and φ.

 

I can't see how it's scientifically justifiable to use a combined signal in this way. The magnitude could remain fairly constant as θ and φ fluctuate wildly, and the filter would never know. I don't know that much about this kind of filtering.

 

No matter how the three signals would combine into one, there's two dimensions that are being "lost".