Human Visual Tracking Research Articles

Global Instance Tracking: Locating Target More Like Humans.

Target tracking, the essential ability of the human visual system, has been simulated by computer vision tasks. However, existing trackers perform well in austere experimental environments but fail in challenges like occlusion and fast motion. The massive gap indicates that researches only measure tracking performance rather than intelligence. How to scientifically judge the intelligence level of trackers? Distinct from decision-making problems, lacking three requirements (a challenging task, a fair environment, and a scientific evaluation procedure) makes it strenuous to answer the question. In this article, we first propose the global instance tracking (GIT) task, which is supposed to search an arbitrary user-specified instance in a video without any assumptions about camera or motion consistency, to model the human visual tracking ability. Whereafter, we construct a high-quality and large-scale benchmark VideoCube to create a challenging environment. Finally, we design a scientific evaluation procedure using human capabilities as the baseline to judge tracking intelligence. Additionally, we provide an online platform with toolkit and an updated leaderboard. Although the experimental results indicate a definite gap between trackers and humans, we expect to take a step forward to generate authentic human-like trackers. The database, toolkit, evaluation server, and baseline results are available at http://videocube.aitestunion.com.

Visual Tracking in Development and Aging.

A moving target is visually tracked with a combination of smooth pursuit and saccades. Human visual tracking eye movement develops through early childhood and adolescence, and declines in senescence. However, the knowledge regarding performance changes over the life course is based on data from distinct age groups in isolation using different procedures, and thus is fragmented. We sought to describe the age-dependence of visual tracking performance across a wide age range and compare it to that of simple visuo-manual reaction time. We studied a cross-sectional sample of 143 subjects aged 7–82 years old (37% male). Eye movements were recorded using video-oculography, while subjects viewed a computer screen and tracked a small target moving along a circular trajectory at a constant speed. For simple reaction time (SRT) measures, series of key presses that subjects made in reaction to cue presentation on a computer monitor were recorded using a standard software. The positional precision and smooth pursuit velocity gain of visual tracking followed a U-shaped trend over age, with best performances achieved between the ages of 20 and 50 years old. A U-shaped trend was also found for mean reaction time in agreement with the existing literature. Inter-individual variability was evident at any age in both visual tracking and reaction time metrics. Despite the similarity in the overall developmental and aging trend, correlations were not found between visual tracking and reaction time performances after subtracting the effects of age. Furthermore, while a statistically significant difference between the sexes was found for mean SRT in the sample, a similar difference was not found for any of the visual tracking metrics. Therefore, the cognitive constructs and their neural substrates supporting visual tracking and reaction time performances appear largely independent. In summary, age is an important covariate for visual tracking performance, especially for a pediatric population. Since visual tracking performance metrics may provide signatures of abnormal neurological or cognitive states independent of reaction time-based metrics, further understanding of age-dependent variations in normal visual tracking behavior is necessary.

Frequency Responses to Visual Tracking Stimuli May Be Affected by Concussion.

Human visual tracking performance is known to be reduced with an increase of the target's speed and oscillation frequency, but changes in brain states following a concussion may alter these frequency responses. The goal of this study was to characterize and compare frequency-dependent smooth pursuit velocity degradation in normal subjects and patients who had chronic postconcussion symptoms, and also examine cases of acutely concussed patients. Eye movements were recorded while subjects tracked a target that moved along a circular trajectory of 10° radius at 0.33, 0.40, or 0.67 Hz. Performance was characterized by the gain of smooth pursuit velocity, with reduced gain indicating reduced performance. The difference between normal and chronic patient groups in the pattern of decrease in the gain of horizontal smooth pursuit velocity as a function of the stimulus frequency reflected patients performing more poorly than normal subjects at 0.4 Hz while both groups performing similarly at 0.33 or 0.67 Hz. The performance of acute patients may represent yet another type of frequency response. The findings suggest that there may be ranges of stimulus frequencies that differentiate the effects of concussion from normal individuals.

The Research of 3D Human Motion Simulation and Video Analysis System Implemented in Sports Training

With the rapid development of computer technology, human motion tracking based on video is a kind of using ordinary camera tracking unmarked human movement technology. It has important application value in automatic monitoring, human-computer interaction, sports analysis and many other fields. This research is a hot research direction in the field of computer vision in recent years. Because of the complexity of the problem and the lack of understanding of the nature of the human visual tracking based on video is always a difficult problem in computer vision. The research content of this article is set in sports training, for motion analysis of non-contact, no interfere with measurement and simulation requirements, the use of computer graphics and computer vision technology, discussing 3D human motion simulation technology based on video analysis.

Detecting eye movement abnormalities from concussion.

An attention-based biomarker may be useful for concussion screening. A key role of attention is to generate time-based expectancies of specific sensory information, and it is postulated that postconcussion cognitive impairments and symptoms may stem from a primary deficit in this predictive timing mechanism. There is a close relationship between gaze and attention, but in addressing predictive timing, there is a need for an appropriate testing paradigm and methods to quantify oculomotor anomalies. We have utilized a continuous predictive visual tracking paradigm because human visual tracking requires predicting the temporal course of a stimulus and dynamically synchronizing the required action with the stimulus. We have shown that concussion patients often show disrupted gaze-target synchronization characterized by large gaze position error variability and overall phase advancement. Various attention components interact with visual tracking, and thus there is a possibility that different neurological and physiological conditions produce identifiable visual tracking characteristics. Analyzing neuromotor functions, specifically oculomotor synchronization, can provide a fast, accurate, and reliable assessment of cognitive functions.

Visual Perception for a Partner Robot Based on Computational Intelligence

We propose computational intelligence for partner robot perception in which the robot requires the capability of visual perception to interact with human beings. Basically, robots should conduct moving object extraction, clustering, and classification for visual perception used in interactions with human beings. We propose total human visual tracking by long-term memory, k-means, self-organizing map, and a fuzzy controller is used for movement output. Experimental results show that the partner robot can conduct the human visual tracking.

Dependence of visual tracking capability upon stimulus predictability

Variable, narrow bandwidths of random noise were used to determine the ability of the human visual tracking system to maintain fixation on moving visual stimuli of various degrees of predictability. The results indicate that there is a continuous relationship between stimulus predictability and tracking capability; the less predictable the stimulus motion the greater the phase shift between stimulus and response at a given frequency. Thus, a predictive component in the system seems to compensate for the relatively poor performance capabilities of the system operating in the unpredictive mode and permits maximally accurate following with the patterns of head and target movements encountered in real life.