Find Fundamental Frequency to know if it´s a Man or a woman or children

I´m having problems on a script because i can´t find de Fundamental Frequency on my fft().

Basicly i have a .wav sound saying " a e i o u " and the script must determinate if i´m a boy a girl or children.

The script that i have only get the frequency of the heights peak ( [maxval,idx] = max(abs(Y))), but i need the frequency of the first peak to get the FF...

Can someone help me?

Marco.

[Walter Roberson]

Basicly i have a .wav sound saying " a e i o u " and the script must determinate if i´m a boy a girl or children.

That is not something that can be determined by computation based on any measurement known to Science. The most you can do assign a statistical ranking.

[Marco Macedo]

Sorry my friend but this is a project for the school, and that answer didn´t help me, i will put my code in here so you can help me...

    clear all
close all
clc
pkg load signal
Fs = 22050;
[x,Fs]= audioread('Homem.wav');
%sound(x1,Fs);
figure(1);
plot(x,'g')
figure(2)
plot(xcorr(x),'r')
Y = fft(x);
freq = 0:Fs/length(x):Fs/2;
Y = Y(1:length(x)/2+1);
plot(freq,abs(Y))
[maxval,idx] = max(abs(Y));
freq = freq(idx)
%[f_sort,idx2] = sort(Y,'descend') ;
%%pick the first four
%iwant = f_sort(1)
%freq2 = freq(idx2)
%data = abs(Y); # Positive values
%[pks idx2] = findpeaks(data);
%[pks locs] = findpeaks(data,Fs,'Threshold',1);
%freq(idx3)

if (freq>80 && freq<150)
m = disp("O individuo e um homem");
endif 
if (freq>150 && freq<250)
m = disp("O individuo e uma Mulher");
endif
if (freq>250)
m= disp("O individuo e uma criança");
endif

filename= "teste.txt";
fid = fopen(filename,"w");
fputs (fid,"Homem.wav : ");
fputs (fid,m);
fclose(fid);

[Walter Roberson]

m = disp("O individuo e um homem");

disp() does not return any value: it only displays to the command window.

endif 

MATLAB does not use endif, just end

[Walter Roberson]

pkg load signal

MATLAB does not have pkg

Answer 1

[Walter Roberson]

Your project is not possible to complete using any known science.

Suppose for example that group A has a mean score on something of 10 and one standard deviation is 3. Suppose group B has a mean score on the same item of 15 and one standard deviation is 3. Now suppose a given individual has a score on the item of 14. Which group do they belong to?

The kind of calculation you are doing would say that because they are closer to 15 (14 actual, 15 mean), that they definitely belong to group B, but that would be incorrect.

Normallizing, (14-15)/3 = -1/3, and the zscore of -1/3 is about 0.369 -- that is, more than 36% of the members in group B will have a test score of 14 or below.

Normallizing, (14-10)/3 = 4/3, and the zscore if 4/3 is about 0.909 -- that is more than 9% of the members in group A will have a test score of 14 or above.

It may be roughly 4 times more likely that the individual is a member of group B than of group A, but for example in the USA, that 9% would be over 25 million people that the calculation would be wrong for if it declared that the person was definitely a member of group B.

Vocal analysis cannot tell you whether an individual is a man, a woman, or a child. It can only tell you something about statistics -- and the statistics of the original measurements are often biased, often having already largely excluded the people more likely to make it difficult to calculate the statistics.