symlet will give better result as compared to Haar and DB4.

Hi Pradeep

I would say that the choice would depend on what you want to do. If the purpose is signal representation, or parameter estimation from wavelet representation, you should use something like continuous wavelet transform, or at least undecimated wavelet transform if you don't want too much redundancy. I know nothing about PPG signals but from what I could see they seem quite oscillatory, therefore you should use a wavelet with good frequency localization. For continuous wavelet transform, good choices could be analytic derivatives of gaussians (defined in the fourier domain by f^alpha exp(-f^2) for positive frequencies f, and zero for negative frequencies. This can be easily implemented using fft.

If you want to go for orthogonal or biorthogonal wavelet bases, I agree with Patil's answer, symmlets would be better (because they have nice Fourier localization and are symmetric)

go for coif1 else db2. The oscillatory nature of PPG or most of the biomedical signals best matches with these two mother wavelets. Refer: http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=7359311&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D7359311