The estimation of the human endpoint impedance interacting with a physical environment provides modelling insights both for the field of human movement science and for the design of innovative controllers for collaborative robotics based on physical human-robot interaction. Most of the existing human impedance estimation methods described in the literature rely on controlled environments closer to a laboratory than to an industrial setting. In this paper, a force perturbation method is proposed without any specific requirements on neither the force nor the position trajectories and without additional sensors placed on the human. The method is illustrated through an experimental study on a benchmark interaction task. The task was selected to be sufficiently variable for preventing the use of average trajectories and the perturbations used for the estimations are sufficiently low amplitude and short to not deteriorate the human performance in the task. The obtained impedance model parameters are in adequacy with the ones found in the literature, suggesting the validity of the proposed approach. The low constraints on the experimental settings make it applicable even out of the lab, for example, in humanrobot collaboration in manufacturing environments, to adjust the robot behaviour to changes in the state of the operator (fatigue, stress) or to adjust to a variety of human operators with different interaction strategies.