Until you reduce it to a last layer state it's very difficult to recognize permutation parity. Essentially you either have one of the 22 possible last layer permutations (21 PLLs + skip) or you don't, in which case you have parity.
Btw, there is only one permutation parity, doesn't matter if it's CP or EP. For example, if you had a last layer where only two edges were swapping, say UR and UL, that cube would have PP. However, you could use legal moves, for instance, a T perm, to swap those two edges and swap two corners (URF UBR) and then you'd still have PP, though it would look like you switched from having edge PP to corner PP. This is obviously contrasted with corner and edge orientation parity which are entirely separate from each other.