Analysis of accelerometer values using Matlab

I have zero experience using Matlab however will be using it more to pursue a college project.

I am looking to enter accelerometer values in attempt to discover filters. Since I have little experience and I would need to alter my program to meet the requirements, my question is to anyone who may have experimented in similar areas - what would be the best format to output the values for analysis ?

[James Tursa]

We will need a bit more detail. What are the requirements? What analyses will you be doing? Will you be integrating the accelerometer values to get velocity and position? What device is producing the accelerometer values? Is it at rest or moving? Etc.

[Stuart Morrissey]

Thanks for the reply James. To the best of my limited knowledge, the idea is to apply band pass filters to the accelerometer (on a moving smartphone) to determine velocity. Realistically I'm not even looking for the answer to do so but rather what would be the best way for input into Matlab. Lets say the sampling rate is about 20 times a second, there will be a large amount of readings - so would the best way be separating the values with a delimiter or any suggestions ?

Answer 1

[Star Strider]

Assuming you have 3-axis accelerometers sampled at 20 Hz, I would save them to an Excel file or .csv file. Those are relatively easy to read.

Then after you read your files into your MATLAB workspace, do a fft to determine approximately where your signal frequencies stop and the noise begins. Design your bandpass filters to eliminate d-c offset and low-frequency noise on the low end, and high-frequency noise on the high end. (My filter design procedure is here: How to design a lowpass filter for ocean wave data in Matlab?) Then with your relatively ‘clean’ data, you can then perform whatever analyses you want on them.

[Stuart Morrissey]

Appreciate the info !, that pretty much answers everything.

[Star Strider]

My pleasure!

[larissa montenegro]

Hello. I have a question regarding this info. I am working as well with a 3-axis accelerometer with the aim to determinate tremor parameters amplitude and frequency. my sample frequency is 50 HZ.

My data is in csv format. The accelerometer give me 3 column values x,y,z. I already export the three variables (x,y,z) in matlab but I don't know how to work them. For example. You suggested to apply fft to the signal. That means for each variable separately?

Thank you for your time. Cheers

[Star Strider]

The fft function will do the transform on all of them together if you want it to (and I would). The signals and the noise should be easily separable, unless you have broadband ‘impulse’ noise, in which instance you will have to also use a Savitzky-Golay filter. Design your filters with the information the fft provides.

Yours is actually a new Question. Please post it as such if you need further information.