1
$\begingroup$

First of all, I'm sorry for my bad English writing, plus, I'm new to DSP.

Background

We have got motion data of 20 cows. 10 cows are healthy and 10 cows are slightly lame.

Lameness is an abnormal gait or stance of an animal that is the result of dysfunction of the locomotor system... Wikipedia

It is expected to observe a significant difference in locomotion data, recorded from these 20 cows. Data has been collected using an electronic sensor consist of Angular velocity ($\omega: rad/s$) and Linear acceleration ($a: m/s^2$). As it was expected, data is periodic:

enter image description here Vectors axis: X(Red), Y(Green), Z(Blue).
Up: Angular velocity ($rad/s$) over time ($ms$).
Down: Linear acceleration ($m/s^2$) over time ($ms$).


Each period, represents a step of cow.

Question

I want to perform a comparison between these two group of data to find out the significant difference between lame and non-lame cows. My question is what is/are choice approach(es) to perform this comparison? Is there a standard test to choose the approach according to the signal data?

My best hypotheses:

  • Maybe some of primary data properties be useful. (e.g average, standard deviation, extremums, etc.)
  • Try to use Fourier transform of data. But I have no idea about this one. (Please guide me if it's possible).
  • Try to use derivative or integral of time series. For example, maybe the Angular acceleration ($\alpha: rad/s^2$) is preferable.

Thanks in advance.

Update

These are some other plots:

Normal: Normal cow (2)

Lame: Lame cow (1)

Lame: Lame cow (2)

$\endgroup$
  • $\begingroup$ This may be a good machine learning classification problem. It looks like you only posted a plot for a good cow, can you post a plot for a lame cow as well? $\endgroup$ – Dan Boschen Jan 2 at 3:42
  • $\begingroup$ @DanBoschen Sure. btw, there are some differences between plots, due to individual differences. $\endgroup$ – Fadavi Jan 2 at 8:43
  • $\begingroup$ Interesting, thanks. I can't see anything obviously different in the plots to know what to even look for, hence I think a good application as an AI/ machine learning classification problem - hopefully someone with more experience in that can comment. $\endgroup$ – Dan Boschen Jan 2 at 14:09
  • $\begingroup$ This is an incredibly open ended question and it would definitely benefit from editing for focus. The key thing to determine here is: What are "...significant difference between lame and non-lame cows..."? Are you looking to classify lame-non lame? Are you looking to find those waveform features that are different between lame-non lame cows? Are you looking to classify the cows into lame-non-lame or are you looking into analysing these data to determine gait differences and from there infer something about the condition of the cow? $\endgroup$ – A_A Jan 6 at 10:45
  • $\begingroup$ Thanks for your comment. As you say, I'm looking to classify cows in lame ane non-lame groups. $\endgroup$ – Fadavi Jan 6 at 13:30

Browse other questions tagged or ask your own question.