I agree with Lynn. Unless you're experiencing serious space issues, leave it alone. I've had this problem several times myself and usually it's a valid process. Only on rare occasions was it a runaway process that wouldn't release the connections.

If you are having disk or runaway process issues, try killing all the users on the instance (or on the database which is actually using the space). This may take a while as there's rollback which needs to be done. Then try the shrink again.

Another option is to add another data file on another disk, make it the default, and let TempDB finish what it's trying to do.

And, if all else fails, restarting the services really is the only way to go.