i have a static accelerometer data. So i assume that all signal are from noise. So i model it like this,

$$X(k) = X(k-1) + U(k)$$

But I plot the $U(k)$ autocorrelation function seems it is not white noise, autocorr(uk)

So, I model it as $\mathrm{AR}(1)$ like

$$U(k) = a U(k-1) + W(k) $$

And from MATLAB I got a = -0.24 and I get $W(k)$. When I plot $W(k)$ it still not white noise. How can I get variance of white noise from my accelerometer data?

  • Welcome to DSP.SE! I don't understand how you get $U(k)$ from your signal. How do you split your original data into the noise-free signal and the noise itself? – Tendero Oct 11 at 14:43
  • A static acclerometer should measure the specific force which is the sum of inertial acceleration, bias (if these are MEMS-based sensors), and noise. So, what you are essentially observing is bias + noise. – Maxtron Oct 11 at 15:22
  • @Tendero : sorry i'm still new in DSP, so i cant get U(k) from accel data that i model it to X(k) = X(k-1) + U(k), is that U(k) = diff(X(k))? – EM Farih Oct 12 at 8:03
  • @Maxtron : So i cant get variance of noise (purely white noise) from accel data? – EM Farih Oct 12 at 8:04
  • No. Because it is a combination of both bias and sensor noise. These are two different terms. Please look at this tutorial for more details. Link – Maxtron Oct 12 at 14:53

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.