Quantizing and characterizing the variance of hand postures in a novel transformation task

This paper presents a numerical approach using principal component analysis (PCA) to quantize and characterize the variance of hand postures in a novel posture transformation task. Five subjects were tested in two tasks in which a cursor can be moved by varying the hand posture. This was accomplished by weighted linear combination of 14 sensors of a data glove. The first task was to move a cursor on computer screen in one dimension horizontally, by posing various hand postures. To increase the complexity of control, in the second task, subjects were asked to move a cursor on computer screen in two dimensions. Joint angles were measured during the experiment by the data glove. In both tasks subjects participated in multiple trials until they achieved smooth cursor movement trajectories. PCA was performed over the postures obtained during the multiple trials of the two tasks. Across the trials, in both the tasks a gradual decrease in the number of principal components was observed. This implies that the variance in the postures decreases with learning. Additionally this might indicate that through learning, subjects adapted postural synergies (or eigen postures) in this novel geometrical environment. Postural synergies when visualized revealed task specific synergies.