Human hand descriptions and gesture recognition for object manipulation

Abstract

This work focuses on obtaining realistic human hand models that are suitable for manipulation tasks. A 24 degrees of freedom (DoF) kinematic model of the human hand is defined. The model reasonably satisfies realism requirements in simulation and movement. To achieve realism, intra- and inter-finger constraints are obtained. The design of the hand model with 24 DoF is based upon a morphological, physiological and anatomical study of the human hand. The model is used to develop a gesture recognition procedure that uses principal components analysis (PCA) and discriminant functions. Two simplified hand descriptions (nine and six DoF) have been developed in accordance with the constraints obtained previously. The accuracy of the simplified models is almost 5% for the nine DoF hand description and 10% for the six DoF hand description. Finally, some criteria are defined by which to select the hand description best suited to the features of the manipulation task.

Hybrid Mapping for the Assistance of Teleoperated Grasping Tasks

“…Nevertheless, the use of vision in most of the robotic applications requires the addition of special marks on the hand to facilitate the pose identification, and has the disadvantage that visual occlusions are quite frequent during the hand movements. A sensorized glove was also used to identify hand signals, for instance using neural nets and fuzzy rules [7] and a graph matching approach [8], to identify hand poses using Principal Component Analysis and discriminant functions [9], to teleoperate robotic anthropomorphic [10] and non-anthropomorphic [11] hands, and to extract information about how the humans perform grasping actions [12] [13]. The problem was also addressed in the scope of programming by demonstration, for instance using neural nets [14] or a nearest neighbor algorithm and some patterns defined by a training session [15].…”

Simplified human hand models based on grasping analysis

“…The rest of the joints are modelled by revolute joints. Forward kinematics and inverse kinematics are described in more detail in S. Cobos et al [3].…”

“…The kinematic analysis of the human hand is focused in the role played by the behaviour of the grip in order to decide the more adequate simplified hand model for a particular manipulation. Principal Component Analysis (PCA) has been used previously on hand poses such as [3], [4], and [5]. By means Manuscript received March 10, 2010.…”

Gesture Recognition Based on Human Grasping Activities Using PCA-BMU

“…In finalizing the grasping signal, Principal Component Analysis (PCA) is used to reduce the data redundancy. PCA generally functions as to reduce the dimensionality of dataset in which there are a large number of interrelated variables, while maintaining as much as possible in dataset changes [3] [4]. This research paper is structured as follows: Section 2 addresses the literature review of the related researches to the several approaches, applications and problems of recognizing the fingers grasping force signal.…”