entropy in WPT

how to calculate entropy cost for each node in WPT after decomposition?

Answer 1

[Anshuman]

Hello Snehal,

There are several types of entropy that can be used, including Shannon entropy, log energy entropy, and others. The most common is Shannon entropy, which can be calculated as follows:

signal = randn(1, 1024); % Example signal

% Perform Wavelet Packet Decomposition
wpt = wpdec(signal, 3, 'db1'); % Decompose signal into 3 levels using Daubechies wavelet
numNodes = wpt.nnodes;

% Calculate entropy for each node
entropyCosts = zeros(1, numNodes);
for i = 1:numNodes
    % coefficients for the current node
    coeffs = wpcoef(wpt, i);
    
    % Calculate probabilities
    p = abs(coeffs) / sum(abs(coeffs));
    
    % Shannon entropy
    entropyCosts(i) = -sum(p .* log2(p + eps)); % eps to avoid log(0)
end

% Display entropy costs
disp(entropyCosts);

Hope it helps