Robotic systems start to coexist around humans but cannot physically interact as humans do due to the absence of tactile sensitivity across their bodies. Various studies have developed a scalable tactile sensor to grant a body-scale robotic skin, yet many faced drawbacks arising from the rapidly increasing number of sensing elements or a limited sensibility to a wide range of touches. This article proposes a body-scale robotic skin composed of multimodal sensing modules and a multilayered fabric, simultaneously utilizing superresolution and tomographic transducing mechanisms. These mechanisms employ fewer sensing elements across a large area and complement each other in perceiving a wide range of stimuli humans can sense. Their measurements are processed to encode spatiotemporal properties of touch, which are decoded by a trained convolutional neural network to classify the touch modality, while their computational costs are minimized for on-device computation. The robotic skin was demonstrated on a commercial robotic arm and interpreted human touches for tactile communication, suggesting its capability as a body-scale robotic skin for further physical interaction.