Data-driven human grasp movement analysis

Abstract

Humans tend to simplify the space of possible grasps they can perform. Yet, the description of human hand motions is very complex, and methods to reduce this complexity have attracted much attention in the motor control literature. Important implications in robot hand design and programming have also generated a wide interest in the robotics research community. Early studies prevalently used direct analysis methods such as visual inspection to define grasp taxonomies. More recently, analytical methods have been employed to perform grasping data dimensionality reduction. In this paper, we present a methodology to reconcile these two distinct and apparently incompatible approaches under a unified framework: this allows us to obtain a data-generated grasp taxonomy along with low-dimensional representations which could be used for human grasping data classification and posture reconstruction, as well as for simplifying grasp planning algorithms and robotic hands programming.