How to extract features using wavelet?

[dosha_logic]

I have a classification problem, I have many nonstationary signals that I need to extract features from using wavelet packet (WP), my questions is how to extract features using wavelet, I did laready took the energy of wavelet coefficients as features but the performance of the classifier is not good because the feature vectore for each class look like the features vectors for the other classes, so how can I extract good features from wavelet, do I have to use all the levels in wavelet decomposition as I am currently using only the coefficints from the final level of decomposition.

Please help me with some ideas I am totally new in wavelet.

 

[cshyamanand]

Re: Need help in wavelet

hi
for classification and to extract the features better go for obtaining the features such as mean,standard deviation, variance...
this are very easy features and will give u good result with respect to wavelets..
i hope so...as i have read some articels long back regarding this things..
all the best