Ok
This is the problem with tools like this - it can be easy to think that because you have an HDL coder, then you should be able to put your code on an FPGA right? wrong.
Basically, you need to re-engineer the algorithm so that its in a suitable format to go onto an FPGA.
1. It needs to be fixed point
2. It needs to be streaming, rather than frame based
3. like it says, no matrices
So if you can fix all that away you go.
I suggest you start by learning a bit about FPGAs and how they work. Its not as simple as using HDL coder from any old model (whatever mathworks want you to think).