Authors:
(1) Rafael Kuffner dos Anjos;
(2) Joao Madeiras Pereira.
Table of Links
2 Related Work and 2.1 Virtual avatars
3.4 Virtual Environment and 3.5 Tasks Description
3.6 Questionnaires and 3.7 Participants
4 Results and Discussion, and 4.1 User preferences
4 RESULTS AND DISCUSSION
In this section, we present the main observations made during the tests as well as the difficulties and suggestions from users about
the test task. To assess the difference between the three user embodied representation both in first and third-person perspective, we collected both objective and subjective data, in the form of logs and inquiries respectively, during the evaluation sessions. For the continuous variable, i.e. time, we used Shapiro-Wilk test to assess data normality. Since all samples were normally distributed, we used the Repeated Measures ANOVA with the Bonferroni correction applied to post-hoc tests for multiple comparisons, and the Paired-Samples T-Test test between two samples, to find statistically significant differences. For discrete data, such as the number of obstacle collisions or responses to the questionnaire, we resorted to the Friedman non-parametric test with Wilcoxon-Signed Ranks post-hoc test, also with a Bonferroni correction.
In the following subsections we present the analysis made based on the results of the questionnaires and log files data obtained during the test.
4.1 User preferences
We made two different comparisons based on the data collected through the questionnaires, between representations on the same perspective and representations between perspectives.
4.1.1 Perspective
When comparing between representations in the First-Person Perspective, we found no statistical differences in the sense of embodiment, easiness to complete the task and fatigue felt during the test. The only exception was on the Q8, where users felt that it was easier to perform the second task (go under the obstacle) with the Abstract representation in comparison with the Point-Cloud representation (Z=-2.64, p<0.05).
When comparing between representations in the Third-Person Perspective, the results were slightly better. Users attributed a higher sense of embodiment to the Point-Cloud Avatar when comparing to the Mesh Avatar, with statistical significance only on the sense of agency (Z=-2.812,p=0<0.01), i.e. sense of control of the virtual body. Even though, users felt more fatigue with the PointCloud avatar against the Mesh Avatar (Z=-2.53,p=0.011), on the Third-Person Perspective.
4.1.2 Representation
On the Mesh Avatars, users felt a stronger sense of embodiment in the First-Person Perspective in all its sub-components: sense of agency (Z=-2.687,p=0.007), body ownership(Z=-2.775,p=0.006) and self-location (Z=-3.574,p<0.005). They also found easier to walk in the virtual environment using 1PP (Z=-3.352,p=0.001). About task easiness, they overall preferred the Mesh Avatar in FirstPerson but with statistical significance only in Tasks 1 (Z=-2.902, p=0.004) and 3(Z=-3.579,p<0.005), Q6 and Q8 respectively.
Point-Cloud Avatars had a similar embodiment feeling in both perspectives, but users found easier to perform the third and fourth tasks, Q8 (Z=-2.771,p=0.006) and Q9 (Z=-3.695,p<0.005) respectively.
The Abstract representation had stronger sense of bodyownership and self-location, but with statistical significance only over self-location (Z=-3.422,p=0.001). Users also found easier to walk around the Virtual Environment (Z=-2.838,p=0.005), avoid obstacles in the First (Z=-2.638,p=0.005) and Third (Z=-3.879,p<0.005) Tasks and on the reaching task (Z=- 2.676,p=0.007).
This paper is available on arxiv under CC BY 4.0 DEED license.