Noise modeling with Markov modeling

[ramesses]

Hi
I'm using accelerometer & horizontal gyroscope in order to replace GPS. Now, I'want to model the noise with first order markov process, to use it in kalman filter.
I recorded measurement on all axes and computed auto-correlation.
This picture represents auto-correlation on one of axes.
http://picpaste.com/pics/autocorrelation_x-qjpnbYJk.1437477728.png

Now, I know that the first order markov process takes the following equation :
w = white noise which has the same variance.
and P is the correlation

My problem is how to fix the value of "P" (know as correlation) ?
thank you

 

[mathman]

Noise is usually modeled as a stationary process, not a Markov process.

 

[ramesses]

Gauss-Markov process ?

 

[mathman]

In general, a noise process has a mean of 0. A Markov process has a mean, at a given time, the value at the last known sample.

 

[ramesses]

The Gauss-Markov process gives a good result as you see in this picture.
Now, I don't understand why I need to use kalman filter in-order to estimate the position's error ?
why we don't integrate directly the Gauss-Markov sequence ?

 

[mathman]

Sorry - Ican't answer your specific questions. I have not worked with the specific process or Kalman filters.