hrv spectral analysis

Hello, I've been trying to perform spectral analysis for hrv for some time already but there all the time problems with it.

I'm using such code:

%Spectral Analysis - WELCH metod

   odstepRR=odstepRR.*1000; %odstepRR it an array with NN intervals

      fs = 400;
      maxF=fs/2;

      odstepRR=detrend(odstepRR,'linear');
      odstepRR=odstepRR-mean(odstepRR);

      %Calculate PSD
      [PSD,F] = pwelch(odstepRR,128,64,[],fs); %uses a hamming window

      VLF = [0.0033 0.04];
      LF = [0.04 0.15];
      HF = [0.15 0.4];

      % find the indexes corresponding to the VLF, LF, and HF bands
      iVLF= (F>=VLF(1)) & (F<=VLF(2));
      iLF = (F>=LF(1)) & (F<=LF(2));
      iHF = (F>=HF(1)) & (F<=HF(2));

      % calculate areas, within the freq bands (ms^2)
      aVLF=trapz(F(iVLF),PSD(iVLF));
      set(handles.vlf, 'string', num2str(aVLF));
      aLF=trapz(F(iLF),PSD(iLF));
      set(handles.lf, 'string', num2str(LF));
      aHF=trapz(F(iHF),PSD(iHF));
      set(handles.hf, 'string', num2str(aHF));
      aTotal=aVLF+aLF+aHF;
      set(handles.tp, 'string', num2str(aTotal));

%     calculate areas relative to the total area (%)
%     pVLF=(aVLF/aTotal)*100;
%     pLF=(aLF/aTotal)*100;
%     pHF=(aHF/aTotal)*100;

      %calculate LF/HF ratio
      lfhf =aLF/aHF;

axes(handles.axes5);
set(handles.text45, 'visible', 'on')
set(handles.axes5, 'visible', 'on')

%plot area under PSD curve
area(F(:),PSD(:),'FaceColor',[.6 .6 .6]);        

grid on;

What am I doing wrong? The program doesn't count the spectral parameters (TP, HF, LF) and the plot is in different freq bands than it should be (on X axis freq should be 0.0-0.5 Hz). Shall be grateful for any help. ;)

[NICOLE MIN]

hi

i just want to know is there any solution to tour code above?

[NICOLE MIN]

im stuck at calculating aLF onwards . please help

thanks

[Apoorva Gangula]

What is odstepRR here?

I'm doing a project for fetal heart rate analysis and need to do spectral analysis as a part of it

please answer this

Answer 1

[Wayne King]

Hi Olga, I think your problem is that your frequency resolution is too poor.

You are only obtaining PSD estimates at steps of 1.5625 Hz. Look at:

diff(F)

This is based on your sampling frequency and window (segment) length.

Therefore the code:

        VLF = [0.0033 0.04];
        LF = [0.04 0.15];
        HF = [0.15 0.4];      
        iVLF= (F>=VLF(1)) & (F<=VLF(2));
        iLF = (F>=LF(1)) & (F<=LF(2));
        iHF = (F>=HF(1)) & (F<=HF(2));

Is going to return a vector of zeros for iVLF, iLF, and iHF. Look at the output for

nnz(iVLF)
nnz(iLF)
nnz(iHF)

You are trying to obtain power estimates on a grid that is much finer than your frequency resolution.

[Olga]

Maybe it is possibe to change my frequency resolution so that it'd be possible to obtain power estimates on level that I need?

[Wayne King]

You would have to increase the length of the data vector. I have asked you this question before in other similar posts. How many samples are in odstepRR?

[Olga]

The are 6672 bytes in odstepRR, that is 834 samples

[Wayne King]

Olga, unfortunately that is not enough samples to get the kind of frequency resolution you are after.

[Olga]

That's too bad. Thanks a lot for your help ;)

[Bach]

you can increase the number of FFT to improve the resolution. The frequency resolution = fs / nfft.

Answer 2

[Pradeep Kumar R]

hi guys i'm doing a project of fetal heart rate variability analysis i'm finding difficult to find and plot the heart rate of fetal i've extracted the fetal signal but only heart rate part i'm not able to come to a solution please help me with that.i've attached my code please help me out

Answer 3

[KYAW KYAW]

Dear Wayne King,

I am facing problem for optimizing the peak frequency location in frequency domain analysis for normal and low respiratory rate.

Theoretically, the highest peak frequency must occured at 0.16667 Hz but in pratical is 0.164 to 0.1799(mostly 0.176 above).

My frequency resolution in X axis is 0.0078 which is in FFT only and I optimized then I got the wrost results.

The low frequency lower than 0.10 Hz(low respiratory rate) has issue for spectral analysis that is can't get the peak frequency in 0.10 Hz around there.

Please advise me why theoretical and experimental results are different and can't achieve for very low frequency.

In addition, why the peak frequeny occured in FFT is not same as AR Burg in some of the cases.

Thanks and best regards

[Wayne King]

please post a separate question for this.

Answer 4

[Pradeep Kumar R]

hi guys i'm doing a project of fetal heart rate variability analysis i'm finding difficult to find and plot the heart rate of fetal i've extracted the fetal signal but only heart rate part i'm not able to come to a solution please help me with that.i've attached my code please help me out