I have downloaded audio files of different instruments

https://soundscrate.com/instruments.html

Lets see os there any correlation on Instrumnets

Now see how these Instruments in 3D space

Dominant Eigen Values

Instruments in 2D and 3D space

Many instruments have similar characteristics, with many points closely grouped together. However, flute audio samples have distinct characteristics that set them apart from other instruments.

This is the reason why audio source separation is considered a challenging task among modern researchers, as the similarities in characteristics among instruments make it difficult to accurately separate the individual sources.

Activity 2.1

Load Audio Data

fs = 44100;

[y1,~] = audioread("andean-flute-theme-low_fybeqsnu1.mp3");

[y2,~] = audioread("zo_-pripyati-gitara.mp3");

[y3,~] = audioread("cirk-barabannaya-drob-i-marsh_-silachi.mp3");

[y4,~] = audioread("jamaican-piano_mytjcqbd.mp3");

[y5,~] = audioread("violin-trill_mk15cun_.mp3");

[y6,~] = audioread("z_uk-truba.mp3");

Create Samples

n_samples = 1

range = 2*fs

start = 2*fs;

flute = create_sample_data(y1(:,1),n_samples,range,start);

gitar = create_sample_data(y2(:,1),n_samples,range,start);

drum = create_sample_data(y3(:,1),n_samples,range,start);

piano = create_sample_data(y4(:,1),n_samples,range,start);

violin = create_sample_data(y5(:,1),n_samples,range,start);

trumpt = create_sample_data(y6(:,1),n_samples,range,start);

Correlation of Flute with Other Instruments

figure()

scatter(flute(:,1),gitar(:,1),5,'black',"filled")

xlabel('Flute Data')

ylabel('Gitar Data')

title('Flute vs Gitar')

grid on

figure()

scatter(flute(:,1),drum(:,1),5,'blue',"filled")

xlabel('Flute Data')

ylabel('Drum Data')

title('Flute vs Drum')

grid on

figure()

scatter(flute(:,1),piano(:,1),5,'cyan',"filled")

xlabel('Flute Data')

ylabel('Piano Data')

title('Flute vs Piano')

grid on

figure()

scatter(flute(:,1),violin(:,1),5,'magenta',"filled")

xlabel('Flute Data')

ylabel('Violin Data')

title('Flute vs Violin')

grid on

figure()

scatter(flute(:,1),trumpt(:,1),5,'red',"filled")

xlabel('Flute Data')

ylabel('Trumpt Data')

title('Flute vs Trumpt')

grid on

function sample = create_sample_data(data_array,n_samples,range,start)

sample = zeros(range,n_samples);

for i=1:n_samples

range_sample = (i-1)*range+1 : i*range;

range_data = ((i-1)*range+1)+start : (i*range)+start;

sample(range_sample) = data_array(range_data);

end

end

Activity 2.2

Analysis of Instruments in 2D and 3D space

fs = 44100;

[y1,~] = audioread("andean-flute-theme-low_fybeqsnu1.mp3");

[y2,~] = audioread("zo_-pripyati-gitara.mp3");

[y3,~] = audioread("cirk-barabannaya-drob-i-marsh_-silachi.mp3");

[y4,~] = audioread("jamaican-piano_mytjcqbd.mp3");

[y5,~] = audioread("violin-trill_mk15cun_.mp3");

[y6,~] = audioread("z_uk-truba.mp3");

n_samples = 6

range = 1000

start = 2*fs;

flute = create_sample_data(y1(:,1),n_samples,range,start);

gitar = create_sample_data(y2(:,1),n_samples,range,start);

drum = create_sample_data(y3(:,1),n_samples,range,start);

piano = create_sample_data(y4(:,1),n_samples,range,start);

violin = create_sample_data(y5(:,1),n_samples,range,start);

trumpt = create_sample_data(y6(:,1),n_samples,range,start);

PCA

x = [flute,gitar,drum,piano,violin,trumpt];

m = mean(x')';

xmr = x-m;

s = cov(xmr');

[V,D] = eigs(s,10);

ymr = V'*xmr;

y = ymr;

bar(diag(D))

ymr = V'*xmr;

y = ymr;

flute_range = 1:6;

gitar_range = 7:12;

drum_range = 13:18;

piano_range = 19:24;

violin_range = 25:30;

trumpt_range = 31:36;

Visualize in 2D

scatter_size = 50

figure()

scatter(y(1,flute_range),y(2,flute_range),scatter_size,'k',"filled",'DisplayName','Flute')

hold on

scatter(y(1,gitar_range),y(2,gitar_range),scatter_size,'g',"filled",'DisplayName','Gitar')

hold on

scatter(y(1,drum_range),y(2,drum_range),scatter_size,'magenta',"filled",'DisplayName','Drum')

hold on

scatter(y(1,piano_range),y(2,piano_range),scatter_size,'cyan',"filled",'DisplayName','Piano')

hold on

scatter(y(1,violin_range),y(2,violin_range),scatter_size,'r',"filled",'DisplayName','Violin')

hold on

scatter(y(1,trumpt_range),y(2,trumpt_range),scatter_size,'b',"filled",'DisplayName','Trumpt')

grid on

xlabel('Eigen Vector 1')

ylabel('Eigen Vector 2')

legend()

Visualize in 3D

scatter_size = 50

figure()

scatter3(y(1,flute_range),y(2,flute_range),y(3,flute_range),scatter_size,'k',"filled",'DisplayName','Flute')

hold on

scatter3(y(1,gitar_range),y(2,gitar_range),y(3,gitar_range),scatter_size,'g',"filled",'DisplayName','Gitar')

hold on

scatter3(y(1,drum_range),y(2,drum_range),y(3,drum_range),scatter_size,'magenta',"filled",'DisplayName','Drum')

hold on

scatter3(y(1,piano_range),y(2,piano_range),y(3,piano_range),scatter_size,'cyan',"filled",'DisplayName','Piano')

hold on

scatter3(y(1,violin_range),y(2,violin_range),y(3,violin_range),scatter_size,'r',"filled",'DisplayName','Violin')

hold on

scatter3(y(1,trumpt_range),y(2,trumpt_range),y(3,trumpt_range),scatter_size,'b',"filled",'DisplayName','Trumpt')

grid on

xlabel('Eigen Vector 1')

ylabel('Eigen Vector 2')

zlabel('Eigen Vector 3')

legend()

function sample = create_sample_data(data_array,n_samples,range,start)

sample = zeros(range,n_samples);

for i=1:n_samples

range_sample = (i-1)*range+1 : i*range;

range_data = ((i-1)*range+1)+start : (i*range)+start;

sample(range_sample) = data_array(range_data);

end

end

Reference

https://soundscrate.com/instruments.html