This is a good question - high coverage alone won't be sufficient because (at least with Illumina data), you will have alignment issues due to high homology within the region.
This is a tool that a colleague of mine developed. I know it gave some results that looked decent, but I also think that particular PI ultimately decided to focus on non-HLA variants (so, maybe it still didn't match normal variant calls):
Here are some other tool that I have heard of (but I haven't tried):
http://www.biomedcentral.com/1471-2164/12/42 (paper for link above, I believe)
I remember some commerical option was considered at one time, but it turned out more than just NGS reads were needed (I think it needed to use special HLA amplicons and we had targeted sequencing data?). Not sure if that was Oximon.
I also heard that 454 was potentially better suited for HLA typing (due to longer reads), so the vendor might be able to provide soem software suggestions from that end.