While it is fairly simple to bit-bang the data, it would be pretty slow to load the data into the display. The Velleman USB kit has a latency of 20mS per command. To clock in 36bits would take a minimum of 1.44 seconds because each bit would take 2 commands. With basic glue logic, such as x4 CD4511's and associated LED displays, you could do it in apx 0.16 of a second because only a total of 8 commands would be needed.
So basically, where we go from here depends on how fast you want to be able to update the display.