Predicting object size from hand kinematics: a temporal perspective

Abstract

Research on reach-to-grasp movements generally concentrates on kinematics values that are expression of maxima, in particular the maximum aperture of the hand and the peak of wrist velocity. These parameters provide a snapshot description of movement kinematics at a specific time point during reach, i.e., the maximum within a set of value, but do not allow to investigate how hand kinematics gradually conform to target properties. The present study was designed to extend the characterization of object size effects to the temporal domain. Thus, we computed the wrist velocity and the grip aperture throughout reach-to-grasp movements aimed at large versus small objects. To provide a deeper understanding of how joint movements varied over time, we also considered the time course of finger motion relative to hand motion. Results revealed that movement parameters evolved in parallel but at different rates in relation to object size. Furthermore, a classification analysis performed using a Support Vector Machine (SVM) approach showed that kinematic features taken as a group predicted the correct target size well before contact with the object. Interestingly, some kinematics features exhibited a higher ability to discriminate the target size than others did. These findings reinforce our knowledge about the relationship between kinematics and object properties and shed new light on the quantity and quality of information available in the kinematics of a reach-to-grasp movement over time. This might have important implications for our understanding of the action-perception coupling mechanism.

Fig 1. Lateral and frontal view of hand model.

In Panel A, the local frame of reference (F_local) determined by using the markers rad (placed on the radial aspect of the wrist), ind1 (placed on the metacarpal joint of the index finger) and lit1 (placed on the metacarpal joint of the little finger). x-y represents the metacarpal plane, y-z represents the sagittal plane. In Panel B, the finger plane defined as x, y and z components of the thumb – index plane defined as passing through the markers thu0 (placed on the trapezium bone of the thumb), ind3 (placed on the tip of the index finger), and thu4 (placed on the tip of the thumb). Additional markers (not used to compute the variables of interest) were placed on the metacarpal and proximal interphalangeal joints of the thumb, the proximal interphalangeal joint the index finger, and the proximal interphalangeal joint and the tip of the little finger.

Fig 2. Grip aperture over time for both small and large object.

In Panel A, grip aperture (mm) over time for both the small and the large object. Error bars represent standard errors. In Panel B, example from a representative participant of the index – thumb distance over time for both large (red) and small (light blue) object. Each line represents one trial.

Fig 3. X, y, and z-thumb over time for both small and large object.

In Panel A, x- and y-thumb coordinate over time for both small and large object. Error bars represent standard errors. In Panel B, example from a representative participant of a tridimensional representation of the x-, y- and z-thumb coordinate for both large (red) and small (light blue) object. Each line represents one trial.

Fig 4. X- and z- components of finger plane over time for both small and large object.

In Panel A, x and z components of finger plane at 10%, 50% and 100% of the movement time for both small and large object. In Panel B, example from a representative participant of the thumb and index finger tips position during the action unfolding.

Fig 5. Predictability values matrix.

Predictability and discriminability of kinematics variables over time. Graphical representation of F-scores of kinematics variables (heat map) and accuracies (line graph) from classification analysis. Please note that F-group value is a combined F-score of all the variables at a given time interval.