I'm sure that there is someone that can give you a much better suggestion than I can as I don't really do PDF's, but you should probably start by seeing if you can capture text from the PDF using TP.
You will likely need to place the PDF in text selection mode and then just try creating a text check and verify that it works. If it does, that's a good start.
Next, you probably have images in your PDF so try creating and verifing a bitmap check against one of the images. I think that this is where it could get tricky and I will explain why.
If you are using Acrobat it can be configured in different ways to render images. If the check fails, you will likely want to try changing these options and possibly turning them off(in Acrobat). It should then, hopefully, work correctly. I found this information from working on a printing project where some print drivers caused issues. Adobe had these same issues and it all had to do with rendering.
Now, again hopefully, you should be able to capture text and images which should be all you really need. This should allow you to do baseline comparison using checks and/or string manipulation. If not, hopefully someone else that has more experience with PDFs can throw in some of their experiences.
I don't know if the latest version of Test Partner works better with Adobe than 5.6 or not... I am using 5.6 and I actually save my pdf files as text and then read the text file to get the data.
It is tedious but a few well placed functions should help.