I am wondering about the mathematical implications of concatenating very small windows from different trials. All of the trials are the same size, and are for the same period of time after a stimuli.

In other words, let's say my data has 3 channels, 6 trials, and 10 data points per trial ( size of data [3x10x6]). What I want to build is a new data set with the same 3 channels, but with concatenating the windows of 2 datapoints for all of the trials, and then repeating this for the all of the possible number of windows. In this example it would be 5 windows (size of new data [3x12x5]).

I assume that any frequency analysis will be pointless due to dramatic changes in the signal, but what other implications to the data should I be concerned about? Would it still be reasonable to remove trending or denormalize this new data?

  • 1
    $\begingroup$ What is the purpose of building this new synthetic data set? $\endgroup$ – dcorking Feb 24 '15 at 14:39
  • $\begingroup$ Is to build a dataset that have more information for a very small period in time. Another option could be upsampling but it won't due for the type of analysis that I am doing. $\endgroup$ – tTimoteof Feb 24 '15 at 15:45
  • $\begingroup$ Any Ideas on this topic? $\endgroup$ – tTimoteof Feb 26 '15 at 14:39
  • $\begingroup$ Best to wait a while - this is a new site, and there are not many biomedical engineers here yet. $\endgroup$ – dcorking Feb 26 '15 at 15:02
  • 1
    $\begingroup$ With so few data points per trial, how could you be sure the timeframe matches up and avoid aliasing? From a signal processing standpoint this doesn't sound like a reasonable concatenation for any sort of continuous analysis. $\endgroup$ – Adam Davis Mar 3 '15 at 18:34

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.