Fourier Transformation of a portion of an audio signal is only resulting in a horizontal line, why?

I am attempting to deconstruct / analyze a song to better understand the nature of music. I have been referencing this link ---->https://www.mathworks.com/help/matlab/math/basic-spectral-analysis.html. I am currently trying to identify two seperate singers in the song by there frequency patterns. I have successfully created a time series of a portion of the song where both singers are singing together on a lyrical verse. The trouble is that when I attempt to fourier transform the time series signal into its frequency components, I get a horizontal line across the frequency domain. I should obviously be getting a series of vertical spikes. Attached below is my resulting plot and the code I am running. I believe the problem stems from the power of the DFT and or the frequency range variables, but I can't for the life of me find a solution, after close analysis of my variables. Thank you in advance for the assistance.

Filename: 'C:\Users\nutri\Documents\cMATLAB\Music Analysis Project for 690\volume_bend_stripped.wav'

CompressionMethod: 'Uncompressed'

NumChannels: 2

SampleRate: 44100

TotalSamples: 11842560

Duration: 268.5388 seonds

Title: []

Comment: []

Artist: []

BitsPerSample: 16

%}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% WAVE FILE PORTION
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
folder = 'C:\Users\nutri\Documents\cMATLAB\Music Analysis Project for 690';
baseFileName = 'volume_bend_stripped.wav'; % Include extension
fullFileName = fullfile(folder, baseFileName);
[y, Fs] = audioread(fullFileName); % Read into array
d = audioinfo(fullFileName);
%audiotxt = load("volume_bend_stripped.txt");
% sound(y, Fs); % Play it


n = length(y); % Sample Size / Magnitude
wrkingsize = 11842560 - 7822; % remove zero value samples
% Fs = 44100; % Sample Frequency / Sample per Second
mfs = Fs/2; % Meaningful Frequency
t = (0:n-1)/Fs; % Time Range for Data
dt = 1/Fs; % Time Increments per Sample
F = fft(n); % Discrete Fourier Transform
a = abs(F); % Amplitude of DFT
f = (0:n-1)*(Fs/n); % Frequency Range
pF = (abs(F).^2)/n; % Power of the DFT

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Both singers singing between 40 - 44 seconds
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
f1 = figure('Name','Full Range','NumberTitle','off', 'Color', [0.85 0.85 0.85]);
tiledlayout(2,1)
x0=300;
y0=150;               
width=950;
height=600;
set(gcf,'position',[x0,y0,width,height])

% Distinguish Frequencies between Singers
r = y(1764000:1940400); 
R = length(r); 
u = pow2(nextpow2(R)); % Transform length
ffR = fft(R,u); % Fourier Transform
%fR = (0:R-1)*(Fs/R); % Frequency Range
fR1 = (0:u-1)*(Fs/u); % Frequency Range
pFR = abs(ffR).^2/u; % Power of the DFT
tR = ((dt*(Fs*40)):dt:(dt*(Fs*44))); % Time Range from 40sec to 44sec
tR1 = (0:dt:(R-1)/Fs); % = tR

nexttile 
p4 = plot(tR,r,'m');
xlim([min(tR)-0.1 max(tR)+0.1])
xlabel('time (s)'), ylabel('Amplitude')
grid on
set(gca,'GridColor','k','LineWidth', 1, 'Color', [0.9 0.9 0.9])

nexttile 
p3 = plot(fR1, pFR, 'b');
%xlim([dt*(Fs*40)-1 dt*(Fs*44)+1])
xlabel('Frequency'), ylabel('Power')
grid on
set(gca,'GridColor','k','LineWidth', 1, 'Color', [0.9 0.9 0.9])

Answer 1

[Star Strider]

Consider:

R = length(r); 
u = pow2(nextpow2(R)); % Transform length
ffR = fft(R,u); % Fourier Transform
%fR = (0:R-1)*(Fs/R); % Frequency Range
fR1 = (0:u-1)*(Fs/u); % Frequency Range
pFR = abs(ffR).^2/u; % Power of the DFT

p3 = plot(fR1, pFR, 'b');

Here, ‘R’ is a scalar, so a constant, and since it has no variation with respect to time (all scalars that I’m aware of don’t vary) its Fourier transform is also a constant. The plot of it is correct.

I suspect that you intended to take the fft of something else.

.

[JAMES DYESS]

Again, the follwoing code below is from https://www.mathworks.com/help/matlab/math/basic-spectral-analysis.html

where they take a certain segment of a time series signal, then take the fft of the signal. What I have done is virtually identical. So, I'm struggling to see why my particular code is failing to produce a similair result.

moan = x(2.45e4:3.10e4);

t = 10*(0:1/fs:(length(moan)-1)/fs);

plot(t,moan)

xlabel('Time (seconds)')

ylabel('Amplitude')

xlim([0 t(end)])

m = length(moan); % original sample length

n = pow2(nextpow2(m)); % transform length

y = fft(moan,n); % DFT of signal

f = (0:n-1)*(fs/n)/10;

power = abs(y).^2/n;

plot(f(1:floor(n/2)),power(1:floor(n/2)))

xlabel('Frequency')

ylabel('Power')

[JAMES DYESS]

One difference between my code and the above guideline is that they were using a .au file, while I am using a .wav file. I also have the .txt file, but does it really matter what I use?

[Star Strider]

I completely agree.

The point I was making is that you are calculating the fft of a scalar. The result was a horizontal line with the ‘y’ value of ‘R’.

I assumed that you wanted to calculate the fft of something else, however I had no idea what that was, or if it was provided in the code. (It was not provided as an attached file.) .

[JAMES DYESS]

I must apologize, I just came to understand what you were saying and what I was doing wrong. I was calculating the fft of the scalar R, when I was meant to calculate the signal range ' r '. After replacing R with r, the result is a proper frequency domain. Thanks for the help.

r = y(1764000:1940400);

R = length(r);

u = pow2(nextpow2(R)); % Transform length

ffR = fft(r,u); % Fourier Transform

%fR = (0:R-1)*(Fs/R); % Frequency Range

fR1 = (0:u-1)*(Fs/u); % Frequency Range

pFR = abs(ffR).^2/u; % Power of the DFT

tR = ((dt*(Fs*40)):dt:(dt*(Fs*44))); % Time Range from 40sec to 44sec

tR1 = (0:dt:(R-1)/Fs); % = tR

[Star Strider]

As always, my pleasure!

No worries!