Tactile Signal Reconstruction Assisted by Video Processing

Because of prior knowledge, a person will deduce the strength of the contact force using only visual input. This research proposes a method for cross-modal signal processing to recover tactile information by using visual input. According to our method, visual information supplements tactile information as an auxiliary source. Our paper starts with a group of photos (GOPs) and uses them as input. Second, to identify regions of in-terest(ROI), we employ the low-rank foreground based attention mechanism (LAM). In order to determine contact force in video frames, we prefer to suggest a linear regression convolutional neural network (LRCNN). Results from experiments support our ability to rebuild sentience. Additionally, simulation results as compared to alternative methods of work, our method can save network costs while also increasing the accuracy of material identification.