Machine Learning Data Window Length

Hi All,

I am experimenting with feature extraction techniques on some accelerometer data in order to create a predictive model(using the classification learner app and various ML approaches). I would like some advice on the window size. The sets of data I have are between 10 and 15 seconds long (sampled at 200Hz). However, I would guess that the features should be extracted from a much smaller window than this.. say 1-2 seconds ?

Once i've imported my ~10 seconds worth of data into Matlab, is it possible to specify a set window size in which the data is analysied and features are extracted ?

Thanks in advance.

[Amy]

Once you've imported your sets of ~10 seconds of data, have you tried just partitioning each of those sets into multiple sets that each correspond to 1-2 seconds of time?