Raw data accelerometer to velocity

How convert accelerometer data triaxial and timestamp (file Csv) to velocity ?

Answer 1

[Bjorn Gustavsson]

Acceleration is derivative of velocity, to get velocity from acceleration summation do you must.

[halidou touré]

can you give a practical example please

[Bjorn Gustavsson]

Let's mock something up for you:

w = [5 3 1];
A = [2 1 1/3];
phi = randn(1,3);
t = 0:0.01:10;                
v = sum(A.*sin(t'*w+phi),2); % this is our 1-D velocity.
a = gradient(v,t); % this is what an accelerometer would measure 
v_1 = cumsum(a.*gradient(t')); % First naive integration of the acceleration
v_2 = cumtrapz(t',a);          % Second naive integration of the acceleration
plot(t,v,t,[v_2,v_1])

These worked fine - except the offset because of incorrect initial velocity. These naive methods will be problematic when there is noise in the accelerometer-data.

Adjust and adapt as you see fit. Investigate the effects of different noise-reducing filterings of the accelerometer data.

HTH

[halidou touré]

thank you for your response but I don't follow you, where do these data come from, I need a detailed explanation

[Bjorn Gustavsson]

You have data of acceleration in 3-D with some time-resolution. You have not shared that data even as a figure in an image. Therefore I can only guess. In order to illustrate how to go about integrating accelerometer data I made a very simple one-dimensional case where I mocked-up a time-evolution of a velocity, v, at times between 0 and 10 in steps of 0.01. For this time-varying velocity I calculated an acceleration (like the output from your accelerometer would give for one of the 3 components). In the last 3rd and 2nd last lines I show how to integrate an acceleration-series to get the corresponding velocity-variation with time. In the last line I plot the mock-up-model-velocity and the 2 integrated-from-acceleration-velocities. You could use these lines for separately for the three components of acceleration.

[halidou touré]

The data accelerometer and time stamp

[Bjorn Gustavsson]

Exactly. From your data, you use the time-stamp as t and each accelerometer-component as a in my code-snippet above, then you will get something similar to v_1 and v_2 for each separate velocity-component.

[halidou touré]

Okay,thanks you so much for your support

[halidou touré]

@Bjorn Gustavsson your email please or linkdin profile