Physical Human–Robot Interaction (pHRI) plays a central role in collaborative robotics, where humans and robots share the same workspace and interact through physical contact. In this context, the ability to accurately localize external contacts on the robot structure is essential to enable intuitive command input and ensure safe human–robot interaction. This work investigated the problem of detecting and classifying external contacts on a KUKA LBR iiwa collaborative robotic arm using exclusively intrinsic joint torque sensing, without the use of any external sensing devices. Contact recognition was addressed by partitioning the KUKA robot arm into predefined contact regions. The classification was then performed by exploiting the temporal evolution of the joint’s torque signals recorded during each contact event.