MATLAB Answers

matlab code for musical note recognition based on frequency

69 views (last 30 days)
i am working on a project to identify the musical note based on frequencies using window methods i need help to write the code and understanding it . ihave a code but its to advanced to understand......
close all;
% Note Recognition
%% Note Initialization
mainNames = char('C', 'C#', 'D', 'D#', 'E', 'F', 'F#', 'G', 'G#', 'A', 'A#', 'B');
names = char('A0', 'A#0' ,'B0');
A0 = 27.5;
ind = 4;
for i = 0:87
data(i+1) = A0 * (2^(1/12))^i;
if i>2
a = [mainNames( rem((ind - 4), 12)+1,:) num2str(fix((ind - 4)/12)+1)];
names = char(names, a);
ind = ind + 1;
% Band Initialization
bands(1) = 20;
for i = 1:87
bands(i+1) = +7+(data(i) + data(i+1))/2;
bands(89) = 4500;
fileName = 'c.wav'; % File name
[y, Fs] = audioread(fileName); % Read audio file
y = (y(:,1) + y(:,2))*4; % Decrease 2 channels to 1
%y = (y(:,1));
y(1:2:end) = 0; %from big vector take only odd index values % Do decimation(decreace the no of samples)
frameLength = 4410*2; % 2 Güzel oldu
endPart = frameLength*ceil(length(y)/frameLength); % complete the last frame %ceil is to round off to nearest integer
y(length(y)+1 : endPart) = 0;
f = linspace(1,Fs,frameLength); %creates linearly spaced vector
harmonics = 0;
for i = 1:round(length(y)/frameLength) % For each frame
% Divide audio into frames
frames(i, 1:frameLength) = y( (frameLength*(i-1)+1):(frameLength*i) )';
frame = y( (frameLength*(i-1)+1):(frameLength*i) )';
frame = frame .* hamming(length(frame))'; % Hamming Window
fframe = abs(fft(frame)); % FFT
fp = sum(fframe);
p = sum(abs(frame));
b = true;
if(p < 200 || fp < 1000) % Put a threshold for processing
b = false;
% Bands
for i=1:88
freqBand(i) = mean(fframe( round(bands(i)/(Fs/frameLength) ):round(bands(i+1)/(Fs/frameLength))))^2;
% Plotting
ylim([-1 1])
hold off
% Desicion
m = find(freqBand == max(freqBand(:)));
if(b) disp(names(m,:)); % Print the result
else disp('.'); end
index = 1;
for i = 1:88
if(freqBand(i) > 2000)
harmonics(index) = i;
index = index+1;
this is the code.....

Accepted Answer

Thiago Henrique Gomes Lobato
To really undestand the method, first try to define how exactly you gonna get the frequencies. Your main steps are :
  1. Define your notes and which frequency do they have (Line 5-23)
  2. Read your sound data and define the window size to evaluate the sound (Line 24-33). Here your code has some flaws. Ex: Decimation change the sampling rate, so you can't simply do it to "decrease number of samples" and don't compensate it afterwards. If you want a mono channel from a stereo recording you normally take the mean, not the sum times 4.
  3. For each evaluation window, take the FFT and check which note has the frequency closest to the maximum amplitude (Line 34-86). In your code the FFT is summed in bands so each note has its "own" band and then you can just check the strongest band.
The algorithm is quite simple, but the way it was implemented in your code is indeed not so optimal. I would suggest you to try to implement it yourself in a more clean way based in those main steps. This should also help you to better internalize it.

  1 Comment

varun taliparambe vitel
varun taliparambe vitel on 14 Jun 2020
thank you
i will try implementing myself but i dont know why those formulas used

Sign in to comment.

More Answers (0)