Question

What would I add to the following code to detect the pitch of the audio signal? (use cepstral method)

clear all;
close all;
clc;

info = audiodevinfo;

info.input(1)
info.input(2)
pause(3); % delete it if necessary

clc;

fs = 44.1e3;
noBits = 16;
noChannels = 1;

recordObject = audiorecorder(fs,noBits,noChannels);

disp('Start speaking...');
record(recordObject);
pause(3);
stop(recordObject);
disp('End of Recording.');

x = getaudiodata(recordObject); %takin in as aperiodic signal
noSamples = length(x); %length of vector x "max(size(x))
time = [1:noSamples]/fs; %time ticks

%DFT: X[k]= (sigma x[n]*exp(-j*2*pi*k/Nf*n)
noSamples = length(x2)
noFrequencies = noSamples;
for ct =1:noSamples
X = nan(size(x2));
for k = 1:noFrequencies
X(k) = sum(x2.*exp(-j*2*pi*k/noFrequencies*n));
end

figure;
subplot(2,1,1);
kAsFrequencyIndicies = 1:noFrequencies;
plot(kAsFrequencyIndicies,abs(X));
box off;
subplot(2,1,2);
frequencyTicks = kAsFrequencyIndicies/noFrequencies*fs;
plot(frequencyTicks,abs(X));
box off;

Answer 1

This code is in MATLAB, It records a signal using audiorecorder function.I recorded a signal where i repeatedly kept saying "A". This code extracts pitch and formant using Cepstral Analysis. At the end it prints letter "A" in Notepad.

Here is a code and the screenshot of it:

clc
rec = audiorecorder(1000,16,1);
disp('Start speaking.')
recordblocking(rec, 2);
disp('End of Recording.');
y = getaudiodata(rec);
w=hamming(2000);
c=y.*w;
x=fft(c);
e=log(abs(x));
g=ifft(e);
d=g(1:length(g)/2);
l=[ones(1,25) zeros(1,975)];
M=transpose(d);
f=M.*l;
%subplot(4,1,2);
%plot(f);
p=f(1:25);
%subplot(4,1,3);
%plot(p);
s=fft(p,8000);
%subplot(4,1,4);
%plot(abs(s));
z=s(1:4000);
%subplot(4,1,1);
%plot(real(z));
k=1;
for i=2:length(z)-1;
    if(z(i-1)<z(i)) & (z(i+1)<z(i))
        formant_mag(k)=z(i);
        formant(k)=i;
        k=k+1;
    else
        continue;
    end
end
%plot(real(z));
%hold on
%plot(formant,formant_mag,'r*');
%hold off
n=fft(z);
%plot(abs(n));
l2=[zeros(1,25) ones(1,975)];
p=l2.';
M=real(p.*d);
%plot(q)
[t_val,t_loc]=max(M);
pitch_period=t_loc
pitch_frequency=(1/pitch_period)*1000
b=['a'];
fid=fopen('word2.txt','w');
fprintf(fid,'%c',b);
fclose(fid);
type word2.txt
status=dos('notepad word2.txt')

This code doesn't perform Noise filtering.