How to obtain the same output in spectral descriptors, using frequency and time domain inputs?

All matlab spectral descriptors as spectralSlope, spectralKurtosis,ect have the choice to use a time domain or a frequency domain input, but I'm unable the get the same (even aproximadetly) result, does anyone knows how to solve this?

Example:

clear;
%load an audio
[x,fs]=audioread('example.wav');

%windowing data
win=round(0.025890625*fs);
over=round(0.01875*fs); % 72%
%Obtain the spectral data
[s, f, t]=spectrogram(...
    x,...
    hamming(win),...
    over,...
    win,fs,'yaxis');

slope_normal(:,1)=spectralSlope(x,fs, Window=hamming(win),OverlapLength=over);
slope_spec(:,1)=spectralSlope(s,f);

figure;plot(slope_normal);hold on;plot(slope_sub);

Answer 1

[Brian Hemmat]

%% Compare Time-Domain and Frequency-Domain Inputs

[x,fs] = audioread('Counting-16-44p1-mono-15secs.wav');

% Define window and overlap length

windowLength = round(0.025890625*fs);

overlapLength = round(0.01875*fs);

win = hamming(windowLength);

fftLength = windowLength;

% Calculate the power spectrum the same was as the spectral descriptors

[~,f,~,ps] = spectrogram( ...

x, ...

win, ...

overlapLength, ...

fftLength,fs,'yaxis','power','onesided');

% Slope using time domain input

slope_timeDomainInput = spectralSlope(x,fs, ...

Window=win,OverlapLength=overlapLength,SpectrumType='power');

% Slope using frequency domain input

slope_frequencyDomainInput = spectralSlope(ps,f);

figure(1)

plot(slope_timeDomainInput,'ro');hold on

plot(slope_frequencyDomainInput,'b*');hold off

title("Error = " +rms(slope_timeDomainInput - slope_frequencyDomainInput))