The grasp taxonomy is designed to facilitate the recognition of the human grasp based on observed sensory data. While there has been a significant amount of work on grasp classification (e.g., [1-6]), these classifications are primarily descriptive in nature and serve as a guide for selection of a grasp given task requirements (such as the amount of power versus dexterity).
In our taxonomy (below), the grasp is first dichotomized into volar and non-volar grasp (according to whether there is palmar-object interaction or not). The non-volar grasp is further subdivided to fingertip grasps (if only the fingertips are involved in the grasp) or composite non-volar grasp (if both fingertips and other finger segments are involved).
The non-volar grasp branch of the taxonomy is:
The terms N_0(P_H) and N_1(P_H) indicate the numbers of fingers and finger segments touching the object, respectively. Notice that the classification of the non-volar grasp is done primarily on these two numbers; the further classification to prismatic or disc/circular grasp is done on the basis of error fit to their respective shapes.
The volar grasp branch of the taxonomy is:
While the non-volar grasp can be recognized directly from its branch of taxonomy, volar grasp is not as easy, due to the high degree of contact between the hand and the object. It turns out that using higher-level description of the grasp facilitates the recognition of a volar grasp.
The grasp taxonomy is augmented with higher-level grasp abstraction concepts such as the virtual finger and opposition space. This is done for the following reasons:
As mentioned earlier, the method of grasp recognition is dependent on whether the grasp is volar or non-volar. If the grasp is non-volar, the non-volar branch of the grasp taxonomy acts as a decision tree in which its leaves are the recognized grasp. However, the volar grasp uses a more sophisticated method for its recognition. See section on grasp recognition.