Manipulative Tasks Identification by Learning and Generalizing Hand Motions

In this work, an approach for extracting features among multiple observations towards manipulation tasks recognition is proposed . The diversity of information such as hand motion, fingers flexure and object trajectory are important to represent a manipulation task. By using the relevant features we can generate a general form to represent a specific dataset of manipulation tasks. The hand motion generalization process is obtained and later, given a new observation, the task can be identified.