Error while creating an EDF file

While creating an EDF file in matlab form a matrix, matlab return an error which doesn't make sense

%getting the name from the original file
Name = string(Infos.OriginalName(1:end-4))+ ".edf";
%Extracting the data of the signal in the proper form
sigdata = [traces(1,:)', traces(2,:)', traces(3,:)'];
%getting the sampling frequency of my recordings
fs = 1000;
% t = 0:1/fs:(size(sigdata,1)-1)/fs; I used that to plot the data 
%making the header for the EDF file
hdr = edfheader("EDF+");
hdr.Recording = string(Infos.RecordingDate);
hdr.StartDate = string(datetime(string(Infos.RecordingDate(1:11)),'InputFormat','dd-MMM-yyyy', 'Format','dd.MM.yy'));
hdr.StartTime = string(datetime(string(Infos.RecordingDate(13:end)),'InputFormat', 'HH:mm:ss','Format','HH.mm.ss'));
hdr.NumDataRecords = 1;
hdr.DataRecordDuration = seconds(length(sigdata(:,1))/fs);
hdr.NumSignals = 3;
hdr.SignalLabels = ["Parietal_1" "Parietal_2" "Reference"];
hdr.PhysicalDimensions = repelem("uV",3);

%% this is where my problem is
hdr.PhysicalMin = min(sigdata);
hdr.PhysicalMax = max(sigdata);
hdr.DigitalMin = [-32768 -32768 -32768];
hdr.DigitalMax = [32767 32767 32767];
EEG_signal = edfwrite(Name,hdr,sigdata,'InputSampleType',"physical");

Everything is fine until a run the last line. Matlab give me this error:

Error using signal.internal.edf.write.validateFieldBytes (line 50)

Each value of "PhysicalMin" field must be less than or equal to 8 bytes. See edfheader documentation for more information.

Error in edfwrite/createFile (line 1667)

signal.internal.edf.write.validateFieldBytes(value, fieldIndex, ...

Error in edfwrite (line 493)

[obj, fileInfo] = createFile(obj, filename, hdr,...

I don't understand because my values in the hdr.PhysicalMin and hdr.PhysicalMax are doubles. when I run the "whos" function on the individual value they are 8 bytes each. Any idea how I could solve the problem ?

[Kartheek Akurati]

Hello Maxime Alessandri ,

Would attach Infos structure and traces signal to this question for further investigation ?

Thanks and regards,

Kartheek

[Hassan Farhat]

Hello,

I assume this is a bug in the code

Please use convert to int32 and it will solve your problem.

Regards

Hassan Farhat

[Kartheek Akurati]

Hello Hassan Farhat,

I would like know more about the usage of 32 bits for sample in EDF/EDF+ format. Currently, we are using only 2 bytes (16 bits) for storing a sample as per the data record specification in EDF+ spec.

Please, let me know about your use spec considerations you are using.

Thanks and regards,

Kartheek

[Hassan Farhat]

Hi Kartheek Akurati

It really depends on the amplitude range of your recorded EEG.

int16() is limited between [-32768 & 32767], anynumber outside this range will be threshold to these maximum limits

ex: int32(65000) = 32767.

So if your signal is higher in amplitude, you need to use higher bits (32) (64 bits is not working)

uint32 is limited between [0 & 65535]

Threfore, if you want to stick to 16 bits you need to rescale your inputs to be limited within the above ranges.

Best Regards

Hassan Farhat

[Maxime Alessandri]

Hello,

I'm sorry I never had the oportunity to answer to you.

I found a solution:

My aquisition software record in volts so my hdr.PhysicalMin and hdr.PhysicalMax were like 0.0005 (whatever). So when I turned them in int32 it was just 0.

The way I managed was to scale the signal by 1000000 so that I was in microvolts. Then the int32 worked like a charm.

Thank you all for your help.

Maxime

Answer 1

[Federico Del Gallo]

I'm having the same problem. Did you find something to fix this error?

[Maxime Alessandri]

Hello,

I found the solution, please see my comment above,

Cheers, Maxime