I have time series vibration signal recorded for 10 seconds. I segmented the vibration signal into a number of frames. I need to use an Auto-regressive model for this signal. AR model is used to extract the data from the vibration signal. The goal of an AR model is to estimate the AR coefficient that can fit the original data as much as possible through an optimization process. I am going to classify the vibration signal using the AR coefficient as the features. In order to find the effective classification accuracy, frame analysis has been done. AR model coefficients are extracted from each frame.

My problem is, how can I develop the AR model? How do I determine the order of polynomial from the AR model? And also how do I find the coefficient?

Perhaps, can anyone provide me the Matlab code for this?

  • $\begingroup$ Hai..How did you find the model parameters first of all without setting up the model? I mean "AR model coefficient are extracted from each frame" how did you do that step? $\endgroup$ – Oliver Jun 12 '15 at 3:17
  • $\begingroup$ Hai @Oliver. thanks for commenting. first, i segmented the signal into 10 frame. so from each frame i am going to extract the AR model coefficient. i noticed in Matlab there are built-in function for Ar model that is 'ar'. so i used the ar built-in function to extract the coefficient. mfb=ar(y1,1,'burg','now'); then i used p=getpvec(mfb); to obtain the coeff. but sir i am not very sure whether i am doing it right or not. $\endgroup$ – Hakd Jun 12 '15 at 5:04
  • $\begingroup$ You specified the model order as 1 (second input argument in ar command). So you might have got only one coefficient. Isn't it? $\endgroup$ – Oliver Jun 12 '15 at 7:31
  • $\begingroup$ @Oliver-yes sir, i ve obtained the order 1 with 1 coefficient. but sir, how do i decide on what order is suitable for my model? because i got almost 120 coeff (extracted features) for 1 sample. how do i decide which order is suitable? $\endgroup$ – Hakd Jun 12 '15 at 8:34
  • $\begingroup$ It is not clear what do you mean by feature for one sample. What I understand is that you acquired 10 sec vibration data which is segmented in to many data frames. For each frame you fit an order-1 AR model. Do you have 120 frames so that you obtain 120 coefficients? They are many ways to find model order for a given frame (data samples). Popular one is Akaike information Criterion (AIC). $\endgroup$ – Oliver Jun 12 '15 at 8:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.