This paper is talking about the Wigner function, which is the Fourier transform of the density matrix $\rho(x,x')$ to be a function of $\rho(x,p)$. The density matrix formulation is arguably more fundamental, but the Wigner function is just a bad set of coordinates for the more fundamental density matrix, see my answer to this question: an example of a quantum system for which wigner function transitions to negative values
The density matrix is real valued on diagonal, and only the complex parts of the matrix are off diagonal. But the mystery (if there is one) of complex numbers in quantum mechanics arises in physical circumstances, when pure states diffract. They diffract in the density matrix formulation too, because the pure states are just special density matrices. The arguments in the paper are overly formal and free of content, and this paper need not be read.
To explain further--- the formulas "28" and "34" in the paper express the complex off diagonal density matrix elements with some other notation, as the author states somewhere at the end. This paper just gives a few new symbols for old things.
Why QM is Complex
If you want to get rid of complex numbers in QM, take the real and imaginary parts of the wavefunction. That sounds glib, and it is: the complex nature of amplitudes means that you have states pared up in twos, $|x_r\rangle$ and $|x_i\rangle$, and these two states are indistinguishable, and if $x_r$ turns into $y_i$ with time, then $x_i$ turns into $-y_r$.
The statement that there is a complex structure to quantum mechanics is simply the statement that there is such a doubling of states. To see why this needs to be so, consider the Eigenstates of H. These have to preserve length so they must rotate into each other, so they must come in these pairs (if you insist on using real numbers, unitary time evolution has only complex eigenvalues, so you can't diagonalize it, but you can block-diagoalize it into 2 by 2 blocks, which have this doubling). Now there is a manifest symmetry of multiplying by "i" (where you introduce an operator "i" which takes the top component of each two-by-two block to the bottom component, and the bottom component to minus the top).
This works for all states except possibly the vacuum (because the vacuum has zero energy, it doesn't have to move at all under time evolution, so it doesn't need a partner). The statement is that there is always an "i" symmetry which squares to -1. This "i" commutes with almost everything else, but doesn't commute with time reversal, which switches the sign of i. This shows you that it's not really complex numbers in QM, but the "i" symmetry above, because time reversal doesn't respect i.
I believe this argument appears somewhere in the literature.