I suggest you look at the problem as a 2-band PR-QMF design problem in discrete time. Then you may design any length (support) Binomial QMF / Daubechies wavelet filters (see https://en.wikipedia.org/wiki/Binomial_QMF). Then, you generate continuous-time discrete wavelet transform basis if your application is a discrete-time application. Otherwise, use discrete-time filter bank, accordingly. See https://web.njit.edu/~akansu/PAPERS/ANA-IWS-WAS-ELSEVIER%20PHYSCOM%202010.pdf Hope this helps.

I suggest you should go through Daubechies Wavelets in Matlab. the Taps and vanishing moments are decided by the order of N in the DbN wavelet. e.g.... if you have Db2 wavelet, it has then 2 vanishing moments and if you Db10 Wavelet then it has 10 vanising moments and so on. Haar is the only wavelet having only one vanising moment. for further help you go for Matlab help

You can view this interesting website :

http://www.aticourses.com/sampler/Wavelets%20-%20A%20Conceptual%20Approach.pdf

Thanks everyone for your answers:

What I'm trying to do is to compare the representations of a cardiac sound signal with several kinds of wavelets in terms of NRMSE (root mean squared error), residual decay and number of terms required. I actually use Matching Pursuit to compare time-frequency dictionaries such as Gabor, Chirp-Gabor and MDCT. I'll like to include wavelets (such as Daubechies) in this study. 

Thanks for your suggestions!

I suggest  you to view the stanford university website