When the user is turning the motor you would have to have to measure the RPM of the motor. The H bridge driving the motor would be set in "brake" mode as this is occurring. The RPM would then be proportional to the torque. You would then have to calculate, based on that RPM and previous calibration, how much current should be supplied to the motor to create an equal torque.
Correct me if my physics are wrong here.
I'd avoid using your hand for testing. A simple, reliable method would be to put a spool on the motor shaft with a string wrapped on it and a heavy object on the end of the string. The falling object would provide a consistent torque as it unwound the spool / spun the motor. And of course you can calculate the real torque by measuring the radius of the spool and knowing the weight of the falling object. You can also verify your torque calculation is correct if you can raise that weight at the same speed.