Yeah, it's a big YMMV. I used to interview people a lot and was frequently appalled to hear what my voice actually sounded like. Then people would tell me what a good phone voice I had. But I do think that in the general case, objectively recording what you're actually doing will be much more informative than not.
I've found Skype conversations rather good for this - it has my video as well as the other person's, and thus increases my awareness of a lot of my body language habits. It also means I'm engaged in a conversation, which seems to bring out significantly different body language.
Edit - many apologies to anyone that feels that this discussion was a waste of time.
I just ran across an article (http://techno-anthropology.blogspot.com/2011/04/rough-guide-to-social-skills-for.html) on Hacker News that gives the barest minimum of a guide for social interaction. Unfortunately this isn't the high-quality advice you need to really handle social situations, though it will help with a few of the worst problems.
A few other rules that will help:
On the physical side:
This is a long list, and it isn't even close to complete.
I'm linking to http://lesswrong.com/lw/372/defecting_by_accident_a_flaw_common_to_analytical/ at the suggestion of David Gerard. It has a lot of deeper discussion into why this is worth knowing.