Wavelet coefficients - Shannon entropy

3 次查看（过去 30 天）

Coo Boo 2012-8-10

1
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/45711-wavelet-coefficients-shannon-entropy

Hi to all,

I found some confusing matters in the "wentropy" function. I think that there are some mistakes in the codes related to the 'shannon' and 'logenergy' entropy. For example for the 'shannon' entropy, we have:

 function ent = wentropy(x,t_ent,in3)
 ...
 ...
 case 'shannon'     % in3 not used.
      x = x(x>0).^2;
      ent = -sum(x.*log(eps+x));
 ...
 ...

As it can be seen in the codes, negative coefficients will not be considered in the entropy calculation, even though the coefficients are squared and the related term of log(.) will not be negative. There is a similar code for the 'logenergy' entropy:

 function ent = wentropy(x,t_ent,in3)
 ...
 ...
 case {'energy','log energy','logenergy'}     % in3 not used.
      x = x(x>0).^2;
      ent = sum(log(x));
 ...
 ...

The help of MATLAB did not notice about any operation like that (elimination of negative coefficients in the 'shannon' entropy calculation). I can not find any reference for which the MATLAB has done via these codes. Please give me guidance about that.

Thanks,