The different characteristics of human performance in selecting receding and approaching targets by rotating the head in a 3D virtual environment

doi:10.3724/SP.J.1041.2023.00009

Abstract

Abstract:

In virtual reality (VR), rotating the head to select a moving target is common. A moving target involves two general directions, that is movement toward or away from a user; thus, knowing the characteristics difference between selecting the approaching target (interception movement) and receding target (pursuit movement) is important for designing an efficient user interface.

In this study, 17 participants (7 males; mean age = 22.5 ± 2.5 years) were given an Oculus Rift helmet-mounted display to wear and instructed to complete a task in a VR environment. They were required to position a small opaque sphere (cursor) that appeared randomly on the left or right side of the visual field into a larger half-transparent moving sphere (target) on the other side of the visual field quickly and accurately by rotating their head. The target moved randomly toward or away from the cursor horizontally, which were both at the participants’ eye height, with a depth of 3 meters. The diameter of the cursor was fixed at 4°. The initial movement amplitude (A; distance between the center of the cursor and target; 20° and 40°), the target tolerance (TT; size difference between the target and cursor; 4°, 6°, and 8°), and the target’s moving velocity (V; 0.5 m/s, 1 m/s, 1.5 m/s, and 2 m/s) were varied (Figure 1). The cursor movement paths were divided into three phases for analysis: acceleration, deceleration, and correction (Figure 2).

Results showed that the average total movement time (MT) of selecting receding targets was significantly larger than that of selecting approaching targets (F(1, 16) = 99.64, p < 0.001, η²_p = 0.86) and the two movements performed differently under different factors (Figure 3). A and TT had a similar influence on the total MT for both movements, which increased as A increased (Pursuit: F(1, 16) = 135.61, p < 0.001, η²_p = 0.89; Interception: F(1, 16) = 132.47, p < 0.001, η²_p = 0.89) and TT decreased (Pursuit: F(1.09, 17.40) = 91.91, p < 0.001, η²_p = 0.85, Greenhouse-Geisser correction; Interception: F(1.03, 16.54) = 40.24, p < 0.001, η²_p = 0.72, Greenhouse-Geisser correction). In contrast, V had an inverse effect on the total MT for the two movements. A large V leaded to a long total MT for the pursuit movement (F(1.38, 22.11) = 48.49, p < 0.001, η²_p = 0.75, Greenhouse-Geisser correction), whereas the total MT for the interception movement decreased as V increased (F(1.16, 18.59) = 12.43, p = 0.002, η²_p = 0.44, Greenhouse-Geisser correction). Moreover, the interception movement showed a light U-shaped relationship between the total MT and V, with the lowest point at 1.5 m/s when A was 20° (Figure 3c). The two movements were further compared in the three phases (Figure 4). The outcome showed that the MT of pursuit movement was only longer in the acceleration phase (F(1, 16) =162.32, p < 0.001, η²_p = 0.91) and deceleration phase (F(1, 16) =119.31, p < 0.001, η²_p = 0.88) compared with the interception movement, while the MTs of two movements were very close in the correction phase (F(1, 16) = 0.001, p = 0.974). Moreover, the three factors had different impact on the two movements in three phases. In the acceleration phase, MT increased for the pursuit movement but decreased for the interception movement as A decreased (Pursuit: F(1, 16) = 50.26, p < 0.001, η²_p = 0.76; Interception: F(1, 16) = 90.03, p < 0.001, η²_p = 0.85) and V increased (Pursuit: F(1, 16) = 50.26, p < 0.001, η²_p = 0.76; Interception: F(1, 16) = 90.03, p < 0.001, η²_p = 0.85). In the deceleration phase, although MT was positively related to A (Pursuit: F(1, 16) = 128.83, p < 0.001, η²_p = 0.89; Interception: F(1, 16) = 309.80, p < 0.001, η²_p = 0.95) and negatively related to V (Pursuit: F(1.73, 27.66) = 11.28, p < 0.001, η²_p = 0.41, Greenhouse-Geisser correction; Interception: F(1.76, 28.20) = 195.38, p < 0.001, η²_p = 0.92, Greenhouse-Geisser correction) for both movements, the MT of interception movement increased more quicky as A increased (F(1, 16) = 38.03, p < 0.001, η²_p = 0.70) and decreased more quickly as V increased (F(1.95, 31.11) = 67.77, p < 0.001, η²_p = 0.81) as compared to the pursuit movement. In the correction phase, the TT had a same impact on the MT for both movements (F(1.32, 21.18) = 0.20, p = 0.728, Greenhouse-Geisser correction), which decreased with an increase of TT (F(1.08, 17.25) = 55.46, p < 0.001, η²_p = 0.78, Greenhouse-Geisser correction). Moreover, the increasing V (F(1.30, 20.73) = 25.32, p < 0.001, η²_p = 0.61, Greenhouse-Geisser correction) and A (F(1, 16) =129.81, p < 0.001, η²_p = 0.89) increased the MT for the pursuit movement, while the MT of the interception movement was not affected by both V and A. Based on the findings, a model was proposed to depict the relationship between the total MT and the three factors, which fit the participants’ performance well (Figure 5).

This study showed that the pursuit and interception movements had different characteristics. Selecting a receding target was more difficult than selecting an approaching target via head rotation, and A and V, but not TT, had a different impact on human performance for the two movements. The empirical findings suggested the importance of considering both movements separately when designing a user interface. The model provides a valid method for quantitatively evaluating the characteristics of human performance in selecting moving targets.

Key words: moving target, head-gaze control, movement trajectory, human performance modeling, human-computer interaction

DENG Chenglong, GENG Peng, KUAI Shuguang. (2023). The different characteristics of human performance in selecting receding and approaching targets by rotating the head in a 3D virtual environment. Acta Psychologica Sinica, 55(1), 9-21.

Figures/Tables 5

Figure 1. Experimental setup, stimulus and design.
Note. (a) A participant wore an Oculus Rift VR HMD to complete the task; (b) The experimental virtual scene. The white translucent sphere was the moving target and the yellow sphere was the cursor that need to be positioned into the target area; (c) The schema of selecting a target that moved away from the cursor horizontally (pursuit movement). Three important parameters in the experiment: movement amplitude (A; distance between the center of the cursor and target in the initial position), target tolerance (TT; size difference between the target and cursor) and target’s moving velocity; (d) The schema of selecting a target that moved toward the cursor horizontally (interception movement).

Figure 2. The definition of the three phases of the cursor’s movement process for the pursuit and interception movements.
Note. Examples of the cursor’s velocity profile in the x-axis direction in pursuit and interception movements, each of which included three phases: acceleration, deceleration and correction. The definition criteria for the three phases were adapted from Deng et al.’s (2019) study. The acceleration phase (AP) started from the beginning of the task to the point of cursor’s maximum speed. The division between the deceleration phase (DP) and correction phase (CP) was determined by the point when the total movement speed was less than the midpoint value between the target velocity and the cursor’s maximum velocity in the pursuit movement or when the total movement speed was smaller than half the maximum speed of the cursor in the interception movement and met one of the following three criteria: (1) the cursor velocity changed from greater than the target speed to less than the target speed in the pursuit movement or the cursor velocity changed from positive to negative in the interception movement; (2) the acceleration value of the cursor reached zero when its sign changed from negative to positive; (3) the acceleration in absolute value was smaller than 0.1 times its maximum values and its sign remained negative for some time. The sign of the velocity in the x-axis direction was used in the definition criteria and its positive value was the direction from the initial position of the cursor to the target along the x-axis.

Figure 3. The total movement time changed with various movement amplitudes, target tolerances and target velocities for the pursuit and interception movements.

Figure 4. Movement time in the acceleration, deceleration and correction phases varied with different movement amplitudes, target tolerances and target velocities for the pursuit and interception movements.

Figure 5. Model’s fitting performance for the pursuit and interception movements.
Note. (a) Model’s fitting performance of all participants’ data for the pursuit movement. $ID=V+1.1lo{{g}_{2}}(2A/TT)$; (b) Model’s fitting performance of all participants’ data for the perception movement. $\text{ }\!\!~\!\!\text{ }ID=1/V+1.3lo{{g}_{2}}(2A/TT)$; (c) R-squared distributions of model training and prediction. We first randomly divided participants into a training group which consisted of two thirds of participants (11 participants) and a test group which consisted of the remaining one third of participants (6 participants), and then we developed a model by adopting the data of the training group and used it to predict the movement time of the test group. This process was repeated 1000 times to generate a distribution of R-squared values. The “training” boxplots was the R-squared distribution of the model with training group data. The “prediction” boxplots represented the R-squared distribution of the training model predicting the test data. The lower and upper boundaries of the box represented the first and third quartiles of the distribution and the horizontal line inside the box represented its median. The whiskers above and below the box were the distribution’s maximum and minimum respectively.

[1]	HUANG Xinjie, ZHANG Chi, WAN Huagen, ZHANG Lingcong. Effect of predictability of emotional valence on temporal binding [J]. Acta Psychologica Sinica, 2023, 55(1): 36-44.
[2]	Wu-Changxua,-Li-Huailongb,-Yang-Qunhuib,--Zhang-Kan . HUMAN PERFORMANCE MODELING IN SPATIAL SEGMENTATION CHINESE HANDWRITING RECOGNIZER [J]. , 2003, 35(06): 767-776.
[3]	Wu-Changxu,Yang-Qunhui,Zhang-Kan,Hu-Yongge,Yang-Lei. HUMAN PERFORMANCE MODELING IN TEMPORARY SEGMENTATION CHINESE HANDWRITING RECOGNIZER [J]. , 2003, 35(04): 483-491.