Abstract
Understanding human intentions based on tactile sensing is pivotal for enabling seamless and intuitive human-robot interaction (HRI). By interpreting physical contact, robots can infer human goals, adjust their behavior in real time, and enhance collaboration in shared environments. However, robots cannot fully understand human intent solely from tactile information and often need to rely on other modalities, such as vision. The tactile information generated by finger contact is intrinsically linked to the corresponding motion. Hence, this paper presents a method that estimates human finger motion directly from tactile features obtained during fingertip contact. Motion parameters are represented by finger poses, and tactile features are extracted as contact area, contact position, etc. A linear model is first proposed to describe the relationship between the contact angle and the contact area. To account for the effect of contact forces on the linear relationship, a three-layer structure model (TLSM) of the finger is introduced to characterize the relationship between contact force and contact area. Combining the two models, the contact angle can be estimated. Then, a local pressure center (LPC) algorithm is proposed to determine the contact position of the fingertip. Finally, inverse kinematics is applied to estimate joint angles and reconstruct finger poses. This work developed a method for estimating finger poses via tactile features, which could enhance the perceptual capabilities of robotic systems and be suitable for human-robot interaction.
Funding Information
  • National Natural Science Foundation of China (52475014)

This publication has 47 references indexed in Scilit: