Top-down guided eye movements.

Eye movements (EMs) are an important aspect of human visual behavior. The temporal and space-variant nature of sampling a visual scene requires frequent attentional gaze shifts (saccades) to fixate onto different parts of an image. Fixations are often directed toward the most informative regions in the visual scene. We introduce a model and its simulation that can select such regions based on prior knowledge of similar scenes. Having representations of scenes as a probabilistic combination of regions with certain properties, it is possible to assess the likely contribution of each region in the successive recognition process. Using Bayesian conditional probabilities for each region given the scene category, the model can then predict the informative value of that region and initiate a spatial information-gathering algorithm analogous to an EM saccade to a new fixation.

Spontaneous eye movements during visual imagery reflect the content of the visual scene.

In nine naïve subjects eye movements were recorded while subjects viewed and visualized four irregularly-checkered diagrams. Scanpaths, defined as repetitive sequences of fixations and saccades were found during visual imagery and viewing. Positions of fixations were distributed according to the spatial arrangement of subfeatures in the diagrams. For a particular imagined diagrammatic picture, eye movements were closely correlated with the eye movements recorded while viewing the same picture. Thus eye movements during imagery are not random but reflect the content of the visualized scene. The question is discussed whether scanpath eye movements play a significant functional role in the process of visual imagery.

Model-based supervisory control in telerobotics.

Model-based approaches can be used to confront several of the challenging performance issues in teleoperation. This paper describes a model-based supervisory control technique for telerobotics. A human-machine interface (HMI) was developed for online, interactive task segmentation and planning utilizing a world model of the telerobotic working environment (TRWE). The task model is transferred intermittently over a low bandwidth communication channel for interpretation, planning, and execution of the task segments through the autonomous control capabilities of a telerobot. For the purposes of outlining tasks, a human operator controls a simulation model to generate a "task sequence script" as a sequential list of desired sub-goals for a telerobot. A graphic user interface (GUI) facilitates the development of the task sequence script with viewing perspectives of the graphic display automatically selected as a function of the operational state and model parameters. Also, because the human operator is specifying discrete model set-points of the TRWE, and allowing the autonomous control capabilities of the telerobot to coordinate the actual trajectory between set-points, a provision is made to preview the proposed trajectory for approval or modification before execution. Preliminary results with a manipulator arm remotely controlled via the Internet demonstrate the utility of the model-based supervisory control technique.

String editing analysis of human visual search.

Eye movement (EM) data were recorded for human subjects performing a visual search task in a stereoscopic computer-generated three-dimensional scene. Each experimental run consisted of six presentations: three different object placements on a common background (quasi-natural scene) were used and one of the placements was repeated three additional times. Raw EM searchpath data were linearized, fixation points were defined via a fixation algorithm and, finally, strings of fixation region labels were obtained based upon a priori regionalization schemes. Use of string editing techniques allowed quantitative comparison of the similarity of various searchpaths. Analysis of the similarity of searchpaths for each subject, as well as across subjects, led us to conclude that presentation of repeated object placements caused each subject to develop a partly self-consistent, but idiosyncratic searchpath based upon a spatial model for that placement pattern.

Efficiency of searchpatterns.

Search theory deals with the efficiency of covering searchpatterns; a classical example is systematic row search with sweep spacing dependent upon window size. Human visual search is a function of contrast and size of the targets and of effective visual lobe size, this latter influenced by clutter and foveal load. Human searchpatterns often are optimal covers. Experimental results presented in this paper demonstrate that this is true for instrumental search with display windows controlled by human subjects and quantified with a formula for efficiency. Visual searchpatterns using free eye movements often develop complex irregular sequential searchpatterns.

Computer analysis of eye movements.

An interactive program package for the acquisition, analysis and plotting of human eye movements is introduced. It is shown that the programs described in this paper can be used by scientists in a wide range of disciplines in spite of their different data analysis requirements. An example dealing with smooth pursuit tracking is given.

Level dependent signal flow in the light pupil reflex. II. Phase velocity of responses to sinusoidal light stimuli.

Pupillary responses to sinusoidal light stimuli were measured over a range of light levels and frequencies. The phase lag and equivalent time delay of these responses were reduced in an approximately log-linear fashion with increasing mean light level (slope = -60 ms/log unit). The magnitude of this level dependence is reduced at higher frequencies, and at higher light levels. This nonlinear level dependent signal flow (LDSF) effect is shown to be essentially independent of target distance (accommodative stimulus) which influences pupil size, and of pupil size itself. Thus most of the level dependence probably resides in the afferent path of the light-pupil reflex arc, before the accommodation signal joins the light signal in the Edinger-West-phal nucleus. A systems model is presented to the LDSF effect described here and in the companion papers (Myers and Stark 1993a, b). When parameters of the model are adjusted to fit pupillary responses to transient light stimuli over a range of light levels, the model simulates reduced phase lag in response to increased mean light level, and the reduction in this LDSF effect with increased mean light level or increasing stimulus frequency without further changes in parameters. This latter reduction explains the relatively small level dependence seen in latency data (-34 ms/log unit). These data will be shown (Myers and Stark 1990b) to be commensurate with reduction in pupil cycle time (increased frequency of oscillation) observed in high gain oscillation experiments as mean brightness increases.(ABSTRACT TRUNCATED AT 250 WORDS)

Model control of image processing: pupillometry.

Smart instruments require on-line computers, special purpose hardware, or both. A pupillometer is described that relies only on a general purpose microcomputer with a frame grabber to process infrared video camera pictures of the human eye. An essential feature of the instrument is that a top-down model controls the image processing algorithms. The model generates regions of interest, ROIs, positioned from knowledge of anatomy and optics of the eye and information from previously analyzed frames. Within these adaptively controlled ROIs, fast, run-time algorithms automatically calculate local thresholds, area measurements, moments for centroid position information, and use pyramiding to shorten calculation time for large pupils. Outputs are precise measurements of pupil size and eye position in real time, with adequate bandwidth for most purposes.

Feedforward stabilization in a bimanual unloading task.

When one hand removes a load from the other hand, feedforward motor commands stabilize the position of the unloaded hand. We studied the stabilization of the postural hand using a novel apparatus that allowed unloading at different rates, and unexpected uncoupling of the unloading force from the postural hand. Feedforward stabilization of hand position was observed in all subjects. This stabilization was achieved both by deactivation of postural agonist muscles and by activation of postural antagonist muscles. The neural feedforward command apparently increased with unloading rate. However, the command only partially canceled the interaction torque generated by removing the load, and stabilization became less effective as unloading rate increased.

The interpretation of kernels--an overview.

The kernel identification method is a powerful technique for mathematically representing the dynamic behavior of a nonlinear system. This technique has been applied to a number of physical and physiological systems. An important development which has enhanced the usefulness of the kernel method has been the interpretation of the internal structure of a system by examining the shapes of the higher-degree kernels. Examples of various nonlinear models with known structure are illustrated to show a repertoire of kernel shapes. Variations in parameters of these models result in well-defined changes in the shapes of the kernels. Also, examples are shown of kernels obtained from physiological systems to demonstrate how examination of kernel shapes can lead to accurate predictions of the dynamic behavior of the physiological system. Finally, limitations of the applicable range of the kernel identification method are discussed.

Postural maintenance during movement: simulations of a two joint model.

Voluntary movements of the upper body are accompanied by anticipatory postural adjustments to the lower body in a standing subject. The long-standing hypothesis is that these anticipatory adjustments serve to counteract the perturbation to the body's center of gravity caused by the voluntary arm movement. This paper presents model simulations investigating the possible roles of anticipatory postural activity that accompanies a rapid, upward arm swing. The model incorporates two (idealized) antagonistic muscle pairs controlling the movements of a double-joint system, with a "shoulder joint" between the arm and stiff body links, and an "ankle joint" between the stiff body-leg segment and the ground. Each muscle is represented by a nonlinear viscoelastic element and also includes proprioceptive feedback. Four inputs to the model define the motor control signals for muscle force generation in both the arm and the postural muscle pairs. The neurological component of the model describes consequences of alternate strategies for cocontractions, stretch reflex activity, and anticipatory and synchronous postural activities (or combinations thereof). Simulations with this model show that: (1) none of the postural maintenance schemes considered in these simulations (including varying anticipation) could suppress the initial backward thrust on the body link; (2) the more important destabilizing perturbation is a subsequent forward sway that, left uncountered by postural activity, would eventually leave the body to fall flat on its face; and (3) anticipatory silencing of the postural extensor followed by a brief period of extensor activation (descending control) and synchronous reflex activity (feedback control) appears to be the most likely postural stabilizing strategy that inhibits the continuous forward sway and is consistent with the experimental evidence.

Behaviour space of a stretch reflex model and its implications for the neural control of voluntary movement.

A nonlinear model for the stretch reflex has recently been used to study the interactions between voluntary and reflex controls during fast, targeted movements. The present study explores the topography of a 'behaviour space' generated by computer simulations of this model under various combinations of values for the gain parameters and time constants in the model's feedback loops. In general, we define a behaviour space to be any set of behavioural characteristics of the simulated movement, such as movement time, peak acceleration or peak velocity. The mathematical model can therefore be viewed as an M x N dimensional map from its parameter space N to a behaviour space M. Here, a one-dimensional behaviour space is explored. This provides a method for quantitatively comparing the different control strategies that might be employed by the nervous system for integrating reflex and descending signals during fast, voluntary movements. The results indicate that an optimal strategy will employ proprioceptive feedback as a means of fine-tuning the braking and clamping activities of fast, goal-directed movements and that descending signals are primarily important for initiating the movement and for controlling reciprocal patterns of muscle activity during the end phase of the movement.

Postural maintenance during fast forward bending: a model simulation experiment determines the "reduced trajectory".

A recent article by Crenna et al. (1987) has shown that fast, forward bending movements are accompanied by a backwards motion of the hips and lower limbs. The ongoing research presented in this brief note expands upon the experimental data described by Crenna and colleagues, concerning the postural activities associated with rapid forward bending in standing man. Our primary experimental tool is the computer simulation method, with the standing subject being represented by a double-joint system: the trunk is modeled as a rigid link mechanically coupled (via a "hip" joint) to the lower body link fixed to the ground (via an "ankle" joint). Each of the two joints in this system is independently controlled by a neurological control model for single joint movements, consisting of an idealized pair of antagonistic muscles (flexor and extensor), their common load, and proprioception from the muscle spindles. This model thereby integrates descending commands with proprioceptive feedback in controlling the joint movements. Our early simulation experiments determine a "reduced trajectory", that is, the physical perturbation to the postural system, due to the voluntary movement, in the absence of any stabilizing activities. These simulation experiments clearly show that an important component of the backward movements in the hips and lower limbs during forward bending is indeed due to the mechanical (physical) coupling between the upper and lower body segments and thus not solely a consequence of the anticipatory postural muscle activity. Simulations also predict that any postural activities in the hips and lower limbs should be a two-fold process: first, some preprogrammed, descending control to the lower body would be required to actively enhance the passive, backwards motion (this is consistent with, though not strictly identical to, the hypothesis of Crenna and colleagues); secondly, there must be a subsequent activation in the anterior muscles of the lower body in order to arrest this backwards motion, since otherwise the uncountered momentum would carry the body backward to the floor in less than half a second after the upper body movement has terminated.

Effects of VDT resolution on visual fatigue and readability: an eye movement approach.

The effects of VDT resolution on visual fatigue and readability were studied. Two kinds of displays with different resolutions (1664 x 1200 pixels and 720 x 350 pixels) and fonts were used. In the first experiment, the subjects read from each display for 1 h to induce fatigue. Reading speed and blink rate while reading were measured. Eye movements during visual smooth pursuit tracking tasks were studied before and after reading; quantitative scoring of eye movement performance showed no significant changes. In the second experiment, readability tests with three different character sizes on both displays were conducted and resulting reading eye movements were analysed. For readability of sufficiently large characters, no significant difference between the high and the standard VDT could be detected. However, for very small characters, higher resolution improved readability.

Modeling the neurological control of human movements.

Although computer models have been extensively used in recent years to understand the way physical systems operate and interact, the enormous power of mathematical modeling and computer simulations has been difficult to implement for the benefit of neuroscientists studying the human motor control system. Nevertheless, homeomorphic models are now being used to explain and predict the neural and biomechanical aspects of different human movements. This paper argues for the importance of regarding model simulations as a supplementary approach to traditional methods of experimental investigation by drawing examples from both the experimental and the modeling literature. The discussion focuses on studies of the triphasic control signal for fast, goal-directed movements and on aspects of sampled data control for slow, tracking movements. The aim of this viewpoint article is to promote a more widespread use of modeling and simulation in the field of motor control.

Simulation studies of descending and reflex control of fast movements.

Several neurological control strategies for fast head movements are considered using computer simulations of a stretch reflex model. Each control strategy incorporates a different amount of proprioceptive feedback contributing to braking and/or clamping the movement. The model behavior for each control strategy is qualitatively compared to experimental data that includes the agonist and antagonist EMGs, and the head position, velocity, and acceleration. Significance of the study is discussed with respect to the characteristic tri-phasic EMG pattern for fast voluntary movements and the possible roles that the stretch reflex may have in contributing to this pattern of activation.

Adapted head- and eye-movement responses to added-head inertia.

Adaptation to inertia added to the head was studied in man by mounting masses on a rigidly attached helmet. Two- to ten-fold increases of inertia were thus produced, while an overhead suspension compensated for the weights. Eye and head positions and corresponding velocities were simultaneously recorded during eye-head tracking of a target stepping at 0.2 Hz in the horizontal direction. Without added inertia, fast gaze movements are type III, the accelerated head movement coming early and the resulting VOR truncating the simultaneous eye movement saccade in both amplitude and velocity. Head oscillations are fast and overcompensated by higher gain VOR. With added inertia, the adapted head movement is slowed and delayed. This permits the eye movement saccade to be completed before head movement begins and to escape truncation; the saccade is normal or slightly increased in amplitude. Head oscillations are slow and compensated by normal gain VOR. Either truncation of the saccade or overcompensation of the VOR leads to eye movement and gaze position error that is corrected for by secondary corrective saccades. These same two errors in gaze coordination could explain the cause of the perceived oscillopsia. Oscillopsia, or continual displacement or instability of the visual worlds, is a symptom of breakdown of space constancy, and was prominent and consistent in perceptual reports of our subjects. Adaptation resulting from adding inertia to the head occurred much faster than that induced by adding prisms or lenses.(ABSTRACT TRUNCATED AT 250 WORDS)