I am conducting a research where I am trying to separate heart and lung sounds using wavelet analysis. In this research I have to use different combinations of wavelets, thresholding rule and thresholding method. I used 7 different wavelets, 4 thresholding rules and 2 methods. Then I computed for correlation coefficient to see the best combination but according to the panel single trial is not enough to say that one combination is the best so I tried to use three different sounds. The scenario now is that I have three correlation coefficients for each combinations. However, when I used different sound, It seems that a different combination is better.

can somebody give me a way on how to choose the best combination.