No abstract

Four experiments were performed to study planning and control of pointing movements in human-computer interaction (HCI). We examined how theories and models derived from motor controI research on natural movements apply to HCI, in order to understand why mediated pointing is difficult, and to suggest strategies for improvinginterfaces. An OPTOTRAK sensor and computer system mapped 3-D hand movements to a 2-D display with less than 25 ms lag. Subjects viewed targets on a graphics display and showed evidence for fast corrective processes when target position was perturbed at movement c set, even when gaze was directed away from the target location for the hand. Experiment 4 revealed that even when gain is perturbed during a pointing movement, adaptation is achieved through adjusting kinematic parameters of the same underlying plan.Results indicate that modelling and analysis of performance in mediated pointing must be based on hand actions required for a task, regardless of the size of the display.Pointing performance will be improved by reducing the scale of hand movements; thus future effort should be focused on improving the sensing resolution of input devices for manual interaction. Emerging from this research is a refined two-part model for predicting movement time, providing a more detailed view than Fitts' law. In addition, the model's ... lll parameter space predicts performance changes when hand movements are scaled, and characterises tasks and devices in terms of difficulties in the two movement phases. The experiments provide a model of performance when pointing with an unencumbered hand, a benchmark against which pointing devices can be compared. The measurement and analysis techniques, and the identification of separable direction, amplitude, and width effects may prove useful for predicting target location while a movement is ongoing, and can be applied to the design of an intelligent interface, employing precomputation to respond to object selection.

We address two main issues: the distinction between time-constrained and spatially constrained tasks, and the separable A and W effects on movement time (MT) in spatially- constrained tasks. We consider MT and 3-D kinematic data from human adults pointing to targets in human-computer interaction. These are better fit by Welford's (1968) two- part model, than Fitts' (1954; Fitts & Peterson 1964) ID model. We identify theoretical and practical implications.

The Virtual Hand Lab (VHL) is an an augmented reality envimnment for conducting experiments in human perception and motor performance that involve grasping, manipulation, and other 3D tasks that people perform with their hands. The hardware and software testbed supports both physical and virtual objects, and object behaviors that cm be specified in advance by experimenters. A testbed for conducting experiments must provide visual stimuli that depend on the configuration of the experimental apparatus, on the specific tasks that are being studied, and on the individual characteristics of each subject. Calibration is an important concern and is the subject of this paper. A proper design leads to independent calibration steps that modularize the subsystems that require calibration and explicitly recognize and order the dependencies among them. We describe how the architecture for the VHL was designed to support independent apparatus-specific, experiment-specific, and subject-specific calibrations. The architecture offers benefits for any augmented reality environment by reducing m-calibration times and identifying appropriate modularization in the software that can result in a more robust and efficient implementation.