hello,
in your paper ,you use "zero-padded to extend DFT vector length to 8192 samples" to get high resolution in spectrum, but sometimes it may cause cause problems .And you have mentioned that "Since the spectrum of PPG signals is mostly made up of harmonic components, this should not cause problems, their separation being usually good enough to resolve them". Considering that, the CZT or Zoom-FFT can be used to decrease the operations. Have you try it? if you haven't, i will try to do it and can show the result if you want.
sorry for my poor English.

Can you tell me how to get the results of the test set?
The original web page has become invalid（ http://www.signalprocessingsociety.org/spcup2015/ )
thank you！

Hi,
I want to ask a question for your source code. How can I plot figures in this paper using Matlab?
Is there any code for plotting figures?

There are generally 23 record of PPG data sets of IEEE SPC 2015 in the paper, and I only have 12 here, which are the same as those in the training_data folder in your dataset. I still lack 11 data sets that I can't find online. Do you have the remaining 11 data sets?