BPM analysis in a song

hi i am working in code to calculate in a Musical composition but i have o

 [queen,fs] = audioread('Test_sound.mp3');

vyrez=queen(1:fs*30); 
trhold= 0.01; % / 0.2
x = length(vyrez); 
cas = round(x/fs);



nfft2=2.^nextpow2(x);
otoc=vyrez';
fy=fft(otoc,nfft2);
fy=fy(1:nfft2/2);
fy=fy';
xfft=fs.*(0:nfft2/2-1)/nfft2;

figure
subplot(2,1,1);
plot(vyrez);
title('Original snippet signal from the song ');


cut_off=1.5e3/fs/2;
order=32;

h=fir1(order,cut_off,'low');

fh=fft(h,nfft2);
fh=fh(1:nfft2/2);

mul=fh.*fy;

% subplot(2,1,2);
% plot(abs(mul));

sig=abs(ifft(mul));

subplot(2,1,2);
plot(sig); 
title('Track snippet signal after filters ');

brk=size(fh); % condition of a break in the loop to bypass the error  

 bps=0;
 peak=0;
 temp=0;

but i have problem in two condition

1- condition to prevent multiple beats

2- condition for beat insertion (0.4 optimal threshold)

and any new idea for code or make it shorter or more simplely

[Walter Roberson]

What difficulty are you observing? We do not have your file and you did not say what is wrong.

[Rik]

I recovered the removed content from the Google cache (something which anyone can do). Editing away your question is very rude. Someone spent time reading your question, understanding your issue, figuring out the solution, and writing an answer. Now you repay that kindness by ensuring that the next person with a similar question can't benefit from this answer.

This page is now archived.

Answer 1

[Walter Roberson]

partial correction.

[queen,fs] = audioread('Test_sound.mp3');
if size(queen,1) > fs*30
    vyrez = queen(1:fs*30,1);  % setting of snippet length (do not set many)
else
    vyrez = queen(:,1);
end

Your old code was assuming that you had at least 30 seconds worth of sound, and was doing linear indexing. However what you really had was about 15 3/4 seconds of sound on each of two channels.

trhold= 0.01; % / 0.2
x = length(vyrez); 
cas = round(x/fs);


% sound(vyrez,fs); 

nfft2=2.^nextpow2(x);
otoc=vyrez';
fy=fft(otoc,nfft2);
fy=fy(1:nfft2/2);
fy=fy';
xfft=fs.*(0:nfft2/2-1)/nfft2;

figure
subplot(2,1,1);
plot(vyrez);
title('Original snippet signal from the song');

% subplot(2,1,1);
% plot(xfft,abs(fy/max(fy)));

cut_off=1.5e3/fs/2;
order=32;

h=fir1(order,cut_off,'low');

fh=fft(h,nfft2);
fh=fh(1:nfft2/2);
fh = reshape(fh, size(fy));

Your fh was coming out as a row vector, but fy is a column vector, and that was a problem.

mul=fh.*fy;

% subplot(2,1,2);
% plot(abs(mul));

sig=abs(ifft(mul));

subplot(2,1,2);
plot(sig); 
title('Track snippet signal after filters');

brk=length(fh); % break condition in loop to bypass error

You were using size() with a single input parameter, but you were dealing with a row vector, so you were getting a vector [1 something] . That wasn't good for testing.

 bps=0;
 peak=0;
 temp=0;

 for k=2:length(sig)-1
%      condition for preventing multiple bit insertion 
      if(sig(k)<trhold && trhold<sig(k-1))

You were using sig(k)<trhold<sig(k-1) which means something completely different to MATLAB

          p=0;
          j=k;
          for j= k:k+100
             
                 p=p+real(sig(j));    
          end
          if(p+5<peak)
          temp=0;
          end 
      end
 % condition for beat inclusion (0.4 optimal threshold)
      if(sig(k)>0.4)
          bps=bps+1;

          j=k;
           for j= k+100
               if(j>=brk)
                  break;
              end
                 
           end
      end
 end

 fprintf('There was %d beats in the song snippet %f seconds long \n', bps, cas);

You had the variables reversed for the message.

 min=cas/60;
 bpm=bps/min;
 
 fprintf('BPM = %f [beat/min]\n',bpm);

 
 figure
 subplot(2,1,1);
 plot(xfft,abs(fy/max(fy)));

subplot(2,1,2);
 plot(abs(mul));

[Walter Roberson]

As to the problem with your code:

Look more carefully at your

      if(sig(k)<trhold && trhold<sig(k-1))

test. Inside that test, under certain conditions, you set the value of temp to 0.

However, you never use the value of temp anywhere.

Also your test

     if(sig(k)>0.4)

is outside the threshold test. You are not checking whether the peak is that high: you are just testing whether each sample is that high.

[Walter Roberson]

Consider for example

j = k : min(k+100, brk);
peak = sum(real(sig(j)));

Replaces your for j loop.

[Walter Roberson]

Replace

      if(sig(k)>0.4&&temp==0)   %sig(k)>sig(k-1)&& sig(k)>sig(k+1)&&
          bps=bps+1;
          temp=1;
          peak=0;
          j=k;
           for j= k:k+100
               if(j>=brk)
                  break;
              end
                 peak=peak+real(sig(j));  
           end
      end

with

    if(sig(k)>0.4&&temp==0)   %sig(k)>sig(k-1)&& sig(k)>sig(k+1)&&
          bps=bps+1;
          temp=1;
          j = k : min(k+100, brk);
          peak = sum(real(sig(j)));
      end

Recall that earlier I gave code and said that it replaces your for loop.