CoChat

Classification and Analysis of Human Body Movement Characteristics Associated with Acrophobia Induced by Virtual Reality Scenes of Heights

Open access
Cheng, Xiankai, Bao, Benkun, Cui, Weidong, Liu, Shuai, Zhong, Jun, Cai, Liming, Yang, Hongbo · 2023 · Sensors

The DOI matches its Crossref record; 39 of the 40 references it cites were found in a registry or an index; 1 of the 16 sources its claims cite were checked for support.

Abstract

Acrophobia (fear of heights), a prevalent psychological disorder, elicits profound fear and evokes a range of adverse physiological responses in individuals when exposed to heights, which will lead to a very dangerous state for people in actual heights. In this paper, we explore the behavioral influences in terms of movements in people confronted with virtual reality scenes of extreme heights and develop an acrophobia classification model based on human movement characteristics. To this end, we used wireless miniaturized inertial navigation sensors (WMINS) network to obtain the information of limb movements in the virtual environment. Based on these data, we constructed a series of data feature processing processes, proposed a system model for the classification of acrophobia and non-acrophobia based on human motion feature analysis, and realized the classification recognition of acrophobia and non-acrophobia through the designed integrated learning model. The final accuracy of acrophobia dichotomous classification based on limb motion information reached 94.64%, which has higher accuracy and efficiency compared with other existing research models. Overall, our study demonstrates a strong correlation between people’s mental state during fear of heights and their limb movements at that time.

Overview

CoChat Summary

## TL;DR

The paper builds a virtual-reality height exposure task plus a wearable-sensor pipeline to classify acrophobia from limb motion. Using 43 usable samples and a voting ensemble, it reports up to 94.64% classification accuracy.

## Key findings

  • From 31 recruited subjects, the authors obtained 43 usable samples after data cleaning, with 16 labeled acrophobia and 27 non-acrophobia.
  • Using 88 motion features and leave-one-out cross validation, the designed voting ensemble achieved 88.37% accuracy on the original feature samples; after K-means SMOTE oversampling, accuracy improved to 91%.
  • With feature selection and correlation filtering, the paper reports a best accuracy of 94.63% for the top 80 feature combination, and also states a final acrophobia dichotomous classification accuracy of 94.64% in the abstract.
  • The best voting ensemble performance reported in Table 3 was 96.2% precision, 92.8% recall, and 94.5% F-score; for class-specific results, non-fear-of-heights accuracy was 96.3% and fear-of-heights accuracy was 93.1%.

## Methods

  • The study used a VR high-altitude scenario with 11 tasks grouped into three stages: rest adaptation, board walking, and jumping; participants walked on a rotating plank, retrieved tennis balls, and completed a jump task.
  • Four wireless miniaturized inertial navigation sensors (WMINS) were attached to the left/right wrists and left/right lower tibiae; data were collected at 30 Hz for 100 s.
  • The analysis extracted 29 features per data dimension from time and frequency domains, reduced features using p-values and correlation thresholds, handled imbalance with K-means SMOTE, and evaluated models with leave-one-out cross validation using a hard-voting ensemble of KNN, extra trees, random forest, SVM, and logistic regression.

## Limitations

  • The paper states the sample was small relative to the feature space: 43 samples versus 3480 extracted feature types, requiring feature selection to reduce overfitting.
  • The paper notes class imbalance between acrophobia and non-acrophobia samples, which is why K-means SMOTE was needed.
  • Evident from the text: the sample was limited to adults aged 23–30 and to one VR-induced laboratory paradigm, so the reported model’s generalizability outside this setting is not established.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/s23125482

Cites

39 of 39 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Introduction to wireless miniaturized inertial navigation sensor (WMINS). (a) Outline dimensions of the sensor and the introduction of the whole set of equipment. (b) Schematic diagram of the attachment position of the sensor to the human body. (c) External magnetic charging of the sensor. (i) No charging state, (ii), charging state.
Figure 1 Introduction to wireless miniaturized inertial navigation sensor (WMINS). (a) Outline dimensions of the sensor and the introduction of the whole set of equipment. (b) Schematic diagram of the attachment position of the sensor to the human body. (c) External magnetic charging of the sensor. (i) No charging state, (ii), charging state.CC BY · PubMed Central
Figure 2 The collected motion data of the left wrist and left lower tibia.
Figure 2 The collected motion data of the left wrist and left lower tibia.CC BY · PubMed Central
Figure 3 Introduction to the VR experiment scene. (a) The connection between the VR environment and the actual scene construction. (b) The plank turns left and right.
Figure 3 Introduction to the VR experiment scene. (a) The connection between the VR environment and the actual scene construction. (b) The plank turns left and right.CC BY · PubMed Central
Figure 4 Eleven kinds of task action display under the actual VR acrophobia experiment.
Figure 4 Eleven kinds of task action display under the actual VR acrophobia experiment.CC BY · PubMed Central
Figure 5 The adjusted experimental data composition display.
Figure 5 The adjusted experimental data composition display.CC BY · PubMed Central
Figure 6 The general diagram of the motion data processing model system for scenes of acrophobia. Ni indicates the data length of each of the 11 tasks, nt indicates the data length of each task after task adjustment, where i and t both indicate the number of tasks.
Figure 6 The general diagram of the motion data processing model system for scenes of acrophobia. Ni indicates the data length of each of the 11 tasks, nt indicates the data length of each task after task adjustment, where i and t both indicate the number of tasks.CC BY · PubMed Central
Figure 7 Classification accuracy of each model. (a) Classification accuracy of each model obtained by training the sample data without K-means smote. (b) Comparison of classification accuracy obtained by training the sample data with and without K-means smote.
Figure 7 Classification accuracy of each model. (a) Classification accuracy of each model obtained by training the sample data without K-means smote. (b) Comparison of classification accuracy obtained by training the sample data with and without K-means smote.CC BY · PubMed Central
Figure 8 (a) Classification effect of voting model with different combinations of feature data. (b) Confusion matrix with the optimal combination of feature data.
Figure 8 (a) Classification effect of voting model with different combinations of feature data. (b) Confusion matrix with the optimal combination of feature data.CC BY · PubMed Central
Figure 9 Classification accuracy of each model under Top 80 feature data samples.
Figure 9 Classification accuracy of each model under Top 80 feature data samples.CC BY · PubMed Central
Figure 10 Classification accuracy of feature combinations in voting ensemble models with different correlation coefficient thresholds.
Figure 10 Classification accuracy of feature combinations in voting ensemble models with different correlation coefficient thresholds.CC BY · PubMed Central
Figure 11 Feature types and corresponding p values after feature selection. (a) Type of feature selected at the wrist joint and the corresponding p-value. (b) Type of feature selected at the lower tibiae and the corresponding p-value.
Figure 11 Feature types and corresponding p values after feature selection. (a) Type of feature selected at the wrist joint and the corresponding p-value. (b) Type of feature selected at the lower tibiae and the corresponding p-value.CC BY · PubMed Central

Verification

Claims the paper makes

1 Valid · 0 Not valid · 9 Not checked · 54 Own results and interpretations

22 more claims

Create a free account to see all 64 claims and to verify the 15 references still open.

See all claims — free account
  • It may state two separate claims as one. 29
  • Probably not a substantial claim. 9
  • The quote only partly states this claim. 2

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Citations

Cites

Show all 39 references

Research works printed but not linked: 4

  1. Heimberg, R.G., and Juster, H.R. (1995). Social Phobia: Diagnosis, Assessment, and Treatment, The Guilford Press.
  2. Guo, G., Wang, H., Bell, D., Bi, Y., and Greer, K. (2003). On The Move to Meaningful Internet Systems 2003: CoopIS, DOA, and ODBASE: OTM Confederated International Conferences, CoopIS, DOA, and ODBASE 2003, Catania, Sicily, Italy, November 3–7, 2003. Proceedings, Springer.
  3. Jakkula, 2006, Tutorial on support vector machine (svm), Sch. EECS Wash. State Univ., 37, 3
  4. Powers, D.M. (2020). Evaluation: From Precision, Recall and F-Measure to ROC, Informedness, Markedness and Correlation. arXiv.

Cited by

See all 7

Metadata

Authors
Cheng, X., Bao, B., Cui, W., Liu, S., Zhong, J., Cai, L., & Yang, H.
Affiliations
  • University of Science and Technology of China (CN)
  • Chinese Academy of Sciences (CN)
Corresponding author
Yang, Hongbo
Venue
Sensors
Publisher
MDPI AG
Published
2023-06-10
Volume
23
Issue
12
Pages
5482
ISSN
1424-8220 (ISSN-L)
OpenAlex
W4380263719
Cited by
7
References
40
Citations by year
2026: 2 · 2025: 3 · 2024: 2
FWCI
1.21
Type
journal article
Language
en
Open access
Locations
  • Sensors (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Key R&D Program of China (2022YFC2405603, BE2022064-2)
  • Key R& D Program of Jiangsu Province (2022YFC2405603, BE2022064-2)
Topics
  • Physical Sciences › Computer Science › Human-Computer Interaction › Virtual Reality Applications and Impacts
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
  • Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
Keywords
acrophobia, fear of heights, virtual reality, limb motion, inertial navigation sensors, psychological state, fear response
MeSH terms
Virtual Reality, Fear, Humans, Phobic Disorders, Human Body

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Classification and Analysis of Human Body Movement Characteristics Associated with Acrophobia Induced by Virtual Reality Scenes of Heights. https://app.cochat.ai/papers/10.3390/s23125482
The paper
Cheng, X., Bao, B., Cui, W., Liu, S., Zhong, J., Cai, L., & Yang, H. (2023). Classification and Analysis of Human Body Movement Characteristics Associated with Acrophobia Induced by Virtual Reality Scenes of Heights. Sensors. https://doi.org/10.3390/s23125482

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.