CODEWITHSUNDEEP
matlabhuffman-code
Tamara Caligari
Tamara Caligari

Reputation: 509

Extended Huffman coding in Matlab

Hi I am having trouble encoding a 1000 symbol message with Extended Huffman coding. I have the dictionary already. I just need to encode the message. However, I am not sure how to do this. Any ideas?

I am using Matlab bdw.

Upvotes: 1

Views: 2413

Answers (1)

Tamara Caligari
Tamara Caligari

Reputation: 509

Here is the code I used:

%Extended Huffman 
prob=0.1;
m=4;

%Generating the probabilities

 for i = 1:2^m
   q(i) = 1;
    for j=0: m-1
      b=2^j;
      if bitand(i-1,b)
          q(i)= q(i)*prob;
      else 
          q(i)= q(i)*(1-prob);
     end
   end
end


disp ('Sum of probabilities');
disp (sum(q));

disp('Entropy per symbol');%should be equal to 1
E=sum(q.*log2(1./q));

disp(E/m); 


%huffman 

s=0:2^m-1; %There are 16 symbols from 0000 -> 1111
[dict,avglen] = huffmandict(s,q); %probabilities

I have tried this approach for extended Huffman and the message size did decrease but not by a lot and I don't know if it is a correct approach. The message was first divided into 4 bits and the decimal value obtained was compared to the dictionary. The new encoded message was then obtained:

for j=(0:4:1000-1)
    newcode=message(j+1:j+4); %Dividing the message into 4 bits and saving the     
                              %corresponding decimal values
   array(:,a)=bi2de(newcode);
   a=a+1;
end

 for(f=1:250)
   for(i=1:15)
     if(array(f)==cell2mat((dict(i,1)))) %cell2mat will obtain the value of the cell
     encodedmsg= horzcat(encodedmsg, dict(i,2)); %horzcat will concatenate the array                    with its corresponding codeword
  end
 end
end

Upvotes: 2

Related Questions