Cohen is a general function, which adds a 2D filter (kernel) after WVD to filter out the coherent. Choi-Williams and Cone Shape are 2 special case of Cohen. Choi-Williams uses exponential filter while Cone Shape uses exponential and cone shape in 2 axis separately. You can use any complex filter in the Kernel, but you need to follow the spec as the help file says.
In SPT 6, most functions are in G, so maybe you can see Cohen in the function of Choi-Williams and Cone Shape there. In SPT 7, NI implemented most functions in dll, and fastened the speed by factor 10x for big data set. It seams that the example of Cohen was ignored, given the resource and project scope.
Shie Qian, the father of NI-JTFA, just told me that function is well
explained in his book.
Sorry for such a late reply.