Using loading_MAHNOB, the ECG signal waveform I obtained seems to be wrong.

[daozihappy]

Dear Dr,

Hello, can you help me with the following problem? Thank you very much !

Using "loading_MAHNOB" , I try to get the ECG signal in MAHNOB.

However, the waveform I obtained seems to be wrong, because it is unlike the standard ECG signal waveform (fig2 of the paper publishing MAHNOB database). The signal I obtained and the standard signal are attached.

【the ECG signal waveform I obtained using "loading_MAHNOB"】

【the standard ECG signal waveform】

Note that, when I run "loading_MAHNOB", I change line 69-72 of "loading_MAHNOB", because I can not obtain the raw signal without this change. line 69-72 of "loading_MAHNOB" was changed to:

total_sec = floor(size(data.Record,2) / phys_data.srate);
phys_data.data = zeros(n_chans+1, (total_sec)*phys_data.srate);
for j = 1:n_chans
      raw_data = data.Record(strcmp(electrode_labels{j},electrode_labels_orig), 1:(total_sec) * 
                            phys_data.srate);
 end
 ECG_signal = phys_data.data(33,:)-phys_data.data(34,:);

---

Can you please help me have a look on this? Thank you very much for all your help!
Best regards

[Gijom]

Could you please precise what happen when you do not change the lines ? Why cannot you obtain the raw signal without this change ?

In your code you only update the structure raw_data for each electrode (overwriting the previous assignment) and then you use phys_data to compute ECG. Since phys_data should be 0 I am surprised you even obtain a signal different from 0.

[daozihappy]

Dear Dr, @Gijom @msoley @JoseCalero @Frankkkkk

Thank you so much for your reply.

If I use the line68-72 of the original code "loading_MAHNOB.m", I can not obtain the true length of one trial. Specifically, there are only two sampling points in the phys_data.data(j,:) (the shape of a trial phys_data.data(j,:) is 1×2) instead of the true length of one trial. In other words, .

For example, for the trial 1 of the subject 10, the shape of "data.Record " is 47×29440. However, after using the line68-72 of the original code, the shape of "phys_data.data" becomes 39×2 instead of 39×29440 where 29440 is the true length of the trial.

The reason of this phenomenon is that "triggers(3)-triggers(1)" is equal to 2.

The line68-72 of the original code "loading_MAHNOB.m":

triggers = find(diff(data.Record(47,:))~=0);                                   (line68)
phys_data.data = zeros(n_chans+1,(triggers(3)-triggers(1)));        (line69)
for j = 1:n_chans                                                                         (line70)
       phys_data.data(j,:) =                                                             (line71)
              data.Record(strcmp(electrode_labels{j},electrode_labels_orig),triggers(1)+1:triggers(3));
end                                                                                             (line72)

---

Thus, I change line 69: phys_data.data = zeros(n_chans+1,(triggers(3)-triggers(1)));
to: phys_data.data = zeros(n_chans+1, (total_sec)*phys_data.srate). I found that, in this way, the raw signals can be successfully extracted, since the shape of "phys_data.data" become 39×29440 (rather than 39×2) when the shape of "data.Record " is 47×29440. Here 29440 is the length of one signal of the trial (trial 1 of the subject 10).

Strangely, there is something wrong with the EXG signal waveform using the code after doing the change.

Actually, in the original code, I don't know what "triggers" in line68-71 used for. Also,I do not know whether it is reasonable to do the change.

Could you please help me check whether the change is right and what's wrong with me ?
Thank you very much for all your help.
Best regards,