If you want to duplicate the work in the paper by Wall (et al.), then you have to find out the parameters that BrainVoyager uses. I cannot find that information after spending all of 2 minutes with Google.
You can choose the K and W (peak and width) parameters based on looking at data, or at the models Wall used. Probably the exact values won't matter much in the final results. The r value determines the size of the undershoot, and is usually in the range 0.1 to 0.2. Again, the exact value probably does not matter much.
You can plot your sample waveform with a command like this (here I've set r = 0.2):
3dDeconvolve -num_stimts 1 -polort -1 -nodata 51 0.5 \
-stim_times 1 '1D: 0' 'TWOGAMpw(4,5,0.2,12,7)' \
-x1D stdout: | 1dplot -stdin -THICK -del 0.5
Making plots like this will let you decide quickly if the parameters make sense to you. The synthetic TR used is 0.5, to make the graphs look smooth.