Why the wav or audio is repeating eventhough I record a new voice. please help me I cant hear my new audio it plays the same audio over again.

clc
clear all
close all
warning off
Fs=8000;%Sampling frequency in hertz
ch=1;%Number of channels--2 options--1 (mono) or 2 (stereo)
datatype='uint8';
nbits=16;%8,16,or 24
Nseconds=5;
NumS = 5; 

%% Create a recorder object
recorder=audiorecorder(Fs,nbits,ch);
datatype='uint8';



%% Record user's voice for 5 sec
disp('Get Ready')
pause(3)

% to record audio data from an input device ...
...such as a microphone for processing in MATLAB
recorder=audiorecorder(Fs,nbits,ch);
disp('Start speaking...')


%Record audio to audiorecorder object,...
...hold control until recording completes
recordblocking(recorder,Nseconds);
disp('End of Recording.');
disp('Recorded');
%Store recorded audio signal in numeric array
x=getaudiodata(recorder,datatype);

%%Training Function
f = voiceFeatures(x);
%%Save Users data
user = cellstr(input('Enter the users name:', 's'));

try
    load database
    F=[F;f];
    C=[C;user];
    if numel(F) > 5
        F = F(end-4:end);
        C = C(end-4:end);
    end
    save database
catch 
    F=f;
    C=user;
    save database F C
end


%Write audio file
[filename, pathname] = uiputfile('*.wav', 'File name for output');
if isnumeric(filename); return; end   %user cancel
fullname = fullfile(pathname, filename);
audiowrite(fullname, x, Fs)


%%Training Function
f = voiceFeatures(x);

%%Play users voice
play(recorder);
x= getaudiodata(recorder);
%Store recorded audio signal in numeric array
data=getaudiodata(recorder,datatype);
plot(x)
figure;

disp('done.')
msgbox('Your voice is registered')

[Jan]

Step through your code line by line. In which line is what repeating?

[kim]

the audio I record is repeating like when I run this code and save an mp3/wav the voice I record doesnt play.It plays the same previous audio over and over again like it doesnt play what I recorded after I run this code.

[kim]

clear all
close all
clc
%% Creating a recorder object
recorder = audiorecorder(16000, 8, 1); % 16000 Hz, 8 bits, 1 channels
%% Record the user's voice for 5s
disp("Please say something.")
drawnow(); % update figures and process callbacks
pause(1) % pauses for 1s
recordblocking(recorder, 4); % records the audio to recorder object for 5s.
disp('Recognizing...')
pause(3)


%% Play the user's voice
play(recorder); % plays the audio from the recorder object
data = getaudiodata(recorder); % data is stored in a 8000 x 2 matrix.
% plot(data); % plots the data
% xlabel("Frequency") 
% ylabel("Normalized values between -1 to 1")
% title("Audio sample")

%% Feature Extraction
f_test = voiceFeatures(data)
%% Find the closest feature from database.
load database
D = [];
for i = 1:size(F, 1)
    distance = abs(F(i) - f_test); % calculates the distance between the stored value in the database and the test value.
    D = [D distance]; % adds the distance d to the vector D.
end
%% Smallest distance
smallest_distance = inf; % initially, we take the smallest distance to be infinity.
index = 0; % initially, we take the index to be 0
for i = 1:length(D)
    if D(i) < smallest_distance % if the current distance is smaller than the smallest distance.
        smallest_distance = D(i); % then replace the smallest distance with the current distance.
        index = i; % change the index to the current distance's index.
    end
end
detected_class = C(index);
disp('User Voice is: ');
detected_class

here is my testing part when I record my voice it doesnt match to the save mp3/wav file

[Jan]

I do not understand, what you are asking for. But you can simplify:

D = [];
for i = 1:size(F, 1)
    distance = abs(F(i) - f_test); % calculates the distance between the stored value in the database and the test value.
    D = [D distance]; % adds the distance d to the vector D.
end
%% Smallest distance
smallest_distance = inf; % initially, we take the smallest distance to be infinity.
index = 0; % initially, we take the index to be 0
for i = 1:length(D)
    if D(i) < smallest_distance % if the current distance is smaller than the smallest distance.
        smallest_distance = D(i); % then replace the smallest distance with the current distance.
        index = i; % change the index to the current distance's index.
    end
end

to the single line:

[smallest_distance, index] = min(abs(F - f_test));

Is the reply of voiceFeatures() a scalar?

[kim]

@Jan I figure it out btw for the first code but the second code I sen it doesnt match to what I record in the training part. what should I type? or any changes so that it would match...

[kim]

It doesnt detect whose voice or similar voice in the database

[Jan]

You try to recognize the voice with the features, which have the minimum distance to recorded features. Therefore you have to know, how these features are stored. You extract them by:

f_test = voiceFeatures(data);

Because voiceFeatures() is no command of Matlab's toolbox function (as far as I know), I cannot guess, what it replies. The line:

distance = abs(F(i) - f_test);

looks, like f_test is a scalar. But how can a voice be represented by a scalar value? This does not sound reasonable. Therefore I've asked you, what the output of voiceFeatures() is. You did not answer this question yet. It does not encourage me, if you ignore my questions.

The explanation "the wav or audio is repeating" is confusing only.

[kim]

function [xPitch] = Features(f)
    F= fft(f(:,1));
%%plot(real(F));
    m=max(real(F));
    xPitch=find(real(F)==m,1);
end

this is the voice features

[Jan]

Again a simplification:

function [xPitch] = Features(f)
    F = fft(f(:, 1));
    [~, xPitch] = max(real(F));
end

Okay. So the function "Features" is, what you call "voiceFeatures" in your code?

You expect one single value to allow to recognize a voice. Very optimistic.

Now focus on the current problem: "It doesnt detect whose voice or similar voice in the database". Explaining a problem by describing, what does not happen, is not useful usually. Explain, what is happening. "the wav or audio is repeating" is not clear yet. Maybe this means, that the buzzing noise of the microphone is the dominant frequency. Then the first speaker is recognized, even if he does not say anything.

[kim]

yea right. sorry Im not good at english. My observation is that in the testing part second code is when I speak or say something is to detect whether whos that voice is I store names when I train in the first code. and when I test it eventhough I didnt speak or say something it detected a user which is incorrect like its constant

[kim]

like this eventhough I didnt write hello in my user when I register it

[Jan]

@kim: I've showed you above, how your code can be simplified. Why do you ignore such suggestions?

You display C(index) as output, although C is not created in the code anywhere. Therefore it must be part of the loaded MAT file.

Do not load the contents of a MAT file directly to the workspace, because this provokes such problems frequently. Store the output in a variable instead:

db = load('database.mat');
[smallest_distance, index] = min(abs(db.F - f_test));

Clean up the contents of the database.mat file. It looks like it contains a lot of junk like "NumS" and "Nseconds". Having a bunch of redundant variables increases the confusion level. Files like "colee..wav" and "COLEE.WAV" are a trap also. You do see, that you loose the overview already in a script with 50 lines. So avoid every addition source of troubles: Keep your workspace clean.