Modeling of low severity soft-tissue injury due to unwanted collisions of a robot in collaborative settings is an important aspect to be treated in safe physical Human-Robot Interaction (pHRI). Up to now, safety evaluations for pHRI were mainly conducted by using safety criteria related with impact forces and head accelerations. These indicate severe injury in the robotics context and leave out low severity injury such as contusions and lacerations. However, for the design of an intrinsically safer robot arm, a reliable evaluation of the collision between a human and a robot that is based on skin injury criteria is essential. In this paper, we propose a novel human-robot collision model with and without covering, which is based on the impact stress distribution. The reliability of the proposed collision model is verified by a comparison with various cadaver experiments taken from existing biomechan-ical literature. Since the stress characteristics acting on the human head can be analyzed with this new collision model, the occurrence of certain soft-tissue injury can be estimated. Furthermore, the method serves for selecting the appropriate covering parameters, as e.g. elastic modulus and thickness, by evaluating the chosen skin injury indices.