Your extraction of color is the same as that of audio. The key is to classify all kinds of color codes, which is easy to do. I'll give you a source program to distinguish audio, which I compiled myself. You can imitate it. Almost. The basic principle is to use BP neural network for identification.
%% Clear environment variables
Chromaticity control center
clear
Extraction and normalization of%% training data prediction data
% download four voice signals
Load data 1 c 1
Load data 2 c2
Load data 3 c3
Load data 4 c4
% 4 characteristic signal matrices are combined into one matrix.
data( 1:500，)=c 1( 1:500，)；
data(50 1: 1000，)=c2( 1:500，)；
Data (100 1: 1500,) =c3( 1:500,);
Data (150 1:2000,) =c4( 1:500,);
Percentage of random sort from 1 to 2000
K = rand (1, 2000);
[m，n]= sort(k)；
% input and output data
Input = data (:,2: 25);
output 1 =data(:， 1)；
% Change the output from 1 dimension to 4 dimensions.
For i= 1:2000
Switch output 1(i)
Case 1
output(i，)=[ 1 0 0]；
Case 2
output(i，)=[0 1 0 0]；
Case 3
output(i，:=[0 0 1 0]；
Case 4
output(i，)=[0 0 0 1]；
end
end
% randomly selected 1500 samples as training samples and 500 samples as prediction samples.
input _ train = input(n( 1: 1500)，:)’；
output _ train = output(n( 1: 1500)，:)'；
input _ test = input(n( 150 1:2000)，:)'；
output _ test = output(n( 150 1:2000)，:)'；
Input data standardization percentage
[inputn，inputps]= mapminmax(input _ train)；
%% network structure initialization
innum = 24
midnum = 25
out num = 4；
Weight percentage initialization
w 1=rands(midnum，innum)；
b 1=rands(midnum， 1)；
w2=rands(midnum，out num)；
b2=rands(outnum， 1)；
w2 _ 1 = w2； w2 _ 2 = w2 _ 1；
w 1 _ 1 = w 1； w 1 _ 2 = w 1 _ 1；
b 1 _ 1 = b 1； b 1 _ 2 = b 1 _ 1；
B2 _ 1 = B2； B2 _ 2 = B2 _ 1；
Learning rate%
xite=0. 1
Alpha = 0.01;
%% online training
for ii= 1: 10
e(ii)= 0；
Because I =1:1:1500.
%% network prediction output
x=inputn(:，I)；
% hidden layer output
For j= 1: 1:midnum
I(j)=inputn(:，i)'*w 1(j，)'+b 1(j)；
iout(j)= 1/( 1+exp(-I(j))；
end
% output layer output
yn = w2 ' * Iout '+B2；
%% weight threshold correction
% calculation error
e=output_train(:，I)-yn；
E(ii)= E(ii)+sum(ABS(E))；
% Calculated weight change rate
dw2 = e * Iout
DB2 = e '；
For j= 1: 1:midnum
s = 1/( 1+exp(-I(j)))；
FI(j)= S *( 1-S)；
end
For k= 1: 1:innum
For j= 1: 1:midnum
dw 1(k，j)=FI(j)*x(k)*(e( 1)*w2(j， 1)+e(2)*w2(j，2)+e(3)*w2(j，3)+e(4)*w2(j，4))；
db 1(j)= FI(j)*(e( 1)* w2(j， 1)+e(2)*w2(j，2)+e(3)*w2(j，3)+e(4)*w2(j，4))；
end
end
w 1 = w 1 _ 1+xite * dw 1’+alfa *(w 1 _ 1-w 1 _ 2)；
b 1 = b 1 _ 1+xite * db 1 '+alfa *(b 1 _ 1-b 1 _ 2)；
w2 = w2 _ 1+xite * dw2 '+alfa *(w2 _ 1-w2 _ 2)；
B2 = B2 _ 1+xite * DB2 '+alfa *(B2 _ 1-B2 _ 2)；
w 1 _ 2 = w 1 _ 1； w 1 _ 1 = w 1；
w2 _ 2 = w2 _ 1； w2 _ 1 = w2；
b 1 _ 2 = b 1 _ 1； b 1 _ 1 = b 1；
B2 _ 2 = B2 _ 1； B2 _ 1 = B2；
end
end
%% speech feature signal classification
inputn_test=mapminmax('apply '，input_test，input PS)；
for ii= 1: 1
For i= 1:500% 1500.
% hidden layer output
For j= 1: 1:midnum
I(j)=inputn_test(:，i)'*w 1(j，)'+b 1(j)；
iout(j)= 1/( 1+exp(-I(j))；
end
fore(:，I)= w2 ' * Iout '+B2；
end
end
%% result analysis
% according to the network output to find out what kind of data belongs to.
Because i= 1:500
output_fore(i)=find(fore(:，i)==max(fore(:，I))；
end
%BP network prediction error
error = output _ fore-output 1(n( 150 1:2000))'；
% Draw a classification diagram of the predicted voice category and the actual voice category.
Figure (1)
plot(output_fore，' r ')
continue
plot(output 1(n( 150 1:2000))'，' b ')
Legend ("predicted voice category" and "actual voice category")
% drawing error map
Figure (2)
Drawing (error)
Title ('BP network classification error',' fontsize', 12)
Xlabel ('voice signal',' fontsize', 12)
Ylabel ('classification error',' fontsize', 12)
%print -dtiff -r600 1-4
K = zero (1, 4);
% Find out which category the misjudged classification belongs to.
Because i= 1:500
If the error (I) ~ = 0
[b，c]=max(output_test(:，I))；
Switch c
Case 1
k( 1)= k( 1)+ 1；
Case 2
k(2)= k(2)+ 1；
Case 3
k(3)= k(3)+ 1；
Case 4
k(4)= k(4)+ 1；
end
end
end
% Find individuals and each category.
Kk = zero (1, 4);
Because i= 1:500
[b，c]=max(output_test(:，I))；
Switch c
Case 1
kk( 1)= kk( 1)+ 1；
Case 2
kk(2)= kk(2)+ 1；
Case 3
kk(3)= kk(3)+ 1；
Case 4
kk(4)= kk(4)+ 1；
end
end
Correct rate%
rightridio=(kk-k)。 /kk