The properties of the foot deployed in a bipedal robot that targets the rendering of a human-like dynamic gait are crucial. Firstly, it has to implement a set of mechanical mechanisms/properties that improve the efficiency of the locomotion. Secondly, it has to integrate a sensory system that captures the interaction with the ground with suitable precision. Both systems-the mechanical and the sensory system-have to be integrated as tightly as possible to keep the overall dimensions and weight low. Being the most distal element of the leg, especially the latter is crucial for favorable leg dynamics. Regarding the structural properties, a modern prosthetic foot poses a good solution and has hence been adopted in various bipeds. Their elaborated structures-mostly made from carbon fiber composites-are designed to imitate the mechanisms of the anthropomorphic counterpart. The following presents a concept to estimate the ground interaction based on the intrinsic deformation of a commercially available prosthesis. To measure the deformation, strain gauges are applied to its main structural elements. Using this information, the center of pressure and the normal force acting on it are estimated. The performance of two approaches-linear regression and neural networks-is presented and compared. Finally, the accuracy of the strain-based estimation is evaluated in two experiments and compared to a conventional force/torque sensor (FTS)-based system and a pressure insole. While the presented work is initially motivated by robotics research, it might as well be transferred to the design of a modern actively actuated prosthesis.