ПЕРСОНАЛИЗИРОВАННАЯ VR-КОРРЕКЦИЯ ПСИХОЭМОЦИОНАЛЬНОЙ НАПРЯЖЁННОСТИ И МОНОТОНИИ НА ОСНОВЕ БИОЛОГИЧЕСКОЙ ОБРАТНОЙ СВЯЗИ

УДК 159.91

Abstract

The study presents the rationale behind the development of the psychophysiological framework for individualized correction of two negative functional states via VR technology, namely, psychoemotional tension and monotony. Its aim is to develop and estimate in preliminary terms a closed-loop VR system where the stimulation type would be chosen depending on the individual’s psychotype, and further adaptation based on the user’s physiological state during the intervention. The algorithmic composition consists of psychotype evaluation, acquisition of multiple physiological parameters, pre-processing, estimation of the current mental state, and VR modification. The system’s technical base relies on using BioRadio for measuring EEG, ECG, heart rate, and respiration rates, as well as Unity app running on Meta Quest 3 VR headset. Two different paths for making estimations have been considered: the cardiorespiratory path for detecting tension, and the EEG path for detecting monotony. Preliminary results on closed-loop operation and effects on the relevant indicators have been obtained on adults representing different psychotypes.

Аннотация

В данной статье предложен способ персонализированной коррекции функционального состояния человека средствами виртуальной реальности. Задача исследования состоит в снижении уровня напряжения и монотонии в условиях эмоциональной нагрузки за счет коррекции функционального состояния средствами VR-среды. В методике представляется соотношение оценки психотипа, регистрации биосигналов и их предобработки, нейросетевой оценки состояния и изменения параметров виртуальной среды в реальном режиме времени. Полученные результаты свидетельствуют о перспективности индивидуализированных VR-коррекций, сочетающих типологическую инициализацию и психофизиологическую адаптацию по состоянию в моменте.

Keywords: virtual reality, functional state, psychophysiology, psychotype, biofeedback, psychoemotional tension, monotony.

Ключевые слова: виртуальная реальность, функциональное состояние, психофизиология, психотип, биологическая обратная связь, психоэмоциональное напряжение, монотония.

Introduction

Well-being in the workplace remains an important psychophysiological and social issue. According to the World Health Organization, unsatisfactory work conditions make employees vulnerable to mental disorders, which, along with depression and anxiety, cause loss of productive days amounting to billions per year [1]. From this perspective, approaches helping to reduce stress levels and restore resources seem to be highly important.

Virtual reality has been recognized as an effective non-drug instrument to improve the state of organism, thanks to immersion in controlled environments and reproducibility of interactions. Research into clinical and experimental settings showed that relaxation via VR helped to decrease subjects' stress perception and physiological indices associated with stress response [2]. Furthermore, systematic review showed virtual natural environments to positively influence psychological and psychophysiological variables, especially those regarding mood, stress, and restoration [3].

At the same time, VR effect depends on individuality. While certain virtual environment might be relaxing for one subject, it could be overexciting for another, so personalization of intervention becomes a vital task. This problem has been studied within ESKAL system and Russian researches based on personality typology and virtual psychocorrection [4-7], which proves that personalized scenario is better than an all-inclusive model of intervention. Therefore, apart from creating relaxing VR, the issue lies in how to initiate and adjust intervention's parameters according to the user's reactions.

The present study introduces a solution to the aforementioned problems in the form of closed-loop psychophysiological system. Here, psychotype does not replace ongoing monitoring of biosignals but serves as the initial value for VR session. Biosignal data adjusts VR parameters dynamically during the process. Hypothesis: psychotype-guided initialization and dynamic adjustment based on physiological monitoring results in greater efficacy compared to static VR environment.

Materials and methods

The study utilized the closed-loop system architecture that consists of five steps, namely psychotype assessment, biosignal acquisition, preprocessing, state estimation, and adaptive VR scene generation. The ESKAL technique, which was used in advance before conducting the experiment, constituted the first personalization step [4]. Corresponding to previous studies on person type-based psychocorrection techniques, the result of this step became the starting point in forming an individual profile of the virtual environment with regard to stimulation, vision load, and preferred mode of interaction [5-7].

For real-time physiological parameters measurements, the authors applied the BioRadio system described by its producer as a programmable data acquisition solution designed for experimental and educational use and capable of recording physiological signals in real time [8]. In this research, BioRadio sensors enabled collecting EEG, ECG, heart rate trace, and respiration signals. They were then sent to PC-based bridge software for preprocessing and further inference, while the VR scene was generated using Unity on the Meta Quest 3 hardware platform. The complete closed-loop contour of the proposed system is shown in Figure 1.

Figure 1. Closed-loop personalized VR correction architecture

A short window approach was applied in the signal processing scheme. Tension state was estimated with the use of 10-second cardiorespiratory window. The inputs for this path included respiration signal, ECG signals' dynamics, heart rate plot, and statistics reflecting the short-term variability of the latter. In addition, a special branch for estimating monotony state was considered. Taking into account the gradual nature of developing this mental state and its association with vigilance decrement, but not instant reaction, multiple successive EEG windows were considered in sequence, not independently. The structure of both estimation branches is presented in Figure 2.

Figure 2. Architecture of the tension and monotony estimation models

Two publicly available data sets were employed for off-line building of estimation models. WESAD data set was used for the construction of a model for the tension branch because it contains multi-modal physiological and motion data collected under the condition of three mental states – neutral, stress and amusement [9]. SEED-VIG data set served as the basis for constructing the model for monotony branch as it deals with the problem of vigilance estimation in driving simulation experiment and uses eye tracking measures to characterize vigilance states [10; 11]. The data sources and their role in the study are summarized in Table 1.

Table 1.

Datasets and their role in the study

In the pilot experiment, adults were chosen as subjects due to their diversified psychotype profiles. Participants went through four stages – background registration, state induction, VR correction phase, and a background after session. In the case of a stress-induced state, the result was represented as a combination of the model and PARS score results. In the monotony test condition, evaluation was conducted through model and spectral EEG characteristics. The reason behind using such approach for the evaluation of monotonous states is that monotonous activity corresponds to vigilance oscillations and thus can be measured through EEG analysis [12].

Model output should be considered separately from diagnosis since the latter does not play any role here. Model results have a completely operational meaning – they can act as smoothed control signal for modifying the environment. This detail is extremely important since one of the main purposes of using VR correction systems is providing an environment with the ability to compensate negative functional states.

Results and discussion

The first result of the project is associated with the creation of a technologically sound personalized loop of the VR process. Namely, the process is characterized by the integration of a personality-oriented starting profile and physiological adaptation online. This means that, rather than using a blank preset, the program had an initial set of personalized settings and adapted visual load, audio load, and visual complexity depending on the estimated physiological condition of the user at each moment in time.

The second result can be considered the availability of two distinct modules for assessing two different states and their readiness to be used for monitoring. The tension module scored 0.864 according to F1-macro metric while the monotony module earned a higher F1-macro score – 0.906. Such results suffice to claim that both models can be used as adaptively-controlled functions. It should be noted that the model was not applied as a tool for diagnostics; its operational purpose entailed estimating the user's state to correct the scene. The comparative model-training results are shown in Figure 3.

Figure 3. Model training results

Pilot testing suggested consistent positive dynamics. As for psychoemotional tension, the model score fell from 0.23 before correction to 0.13 after correction. Similarly, the PARS score also dropped from 4.47 to 3.89. There were improvements in this parameter in the majority of subjects. These findings confirm existing research that immersive virtual reality is capable of promoting stress relief specifically if such experiences are designed in an intentional way and not merely serve as visually distracting experiences [2; 3]. The preliminary tension-correction outcomes are visualized in Figure 4.

Figure 4. Preliminary outcomes for tension correction

As far as monotony is concerned, EEG band power analysis validated the outcomes obtained via the model analysis. After correction, it was found that the data showed a drop in signs linked to monotony, such as theta dominance and the imbalance between low-frequency waves and high-frequency waves. This result is quite physiological since vigilance decreases during the completion of monotonous tasks due to changes in the EEG oscillatory dynamics [10-12]. The EEG band-power dynamics are presented in Figure 5, and the integrated monotony reduction indicators are shown in Figure 6.

Figure 5. EEG band-power markers of monotony before and after VR correction

Figure 6. Monotony reduction indicators

Table 2. Summary of preliminary pilot indicators

The results suggest that a combination of psychotype with online physiology leads to a methodologically more robust technique than utilizing either factor separately. Psychotype is valuable in characterizing the first pattern of exposure, yet its stability is such that it does not track the changes during the same session. The main pilot indicators are additionally summarized in Table 2. On the other hand, physiological parameters are reactive to the changing dynamics of the moment and tell little about what kind of stimulation the user prefers before entering the session.

On the other hand, the current results can only be viewed as preliminary. It will be necessary to increase the sample size and move forward in terms of evaluating the effectiveness of this protocol. Instead of presenting the results from the current phase, the next phase should focus on controlled observations and statistical analysis.

Conclusion

In this paper, the psychological and physiological basis of a personalized closed-loop VR correction of psychoemotional tension and monotony is outlined. The suggested approach combines the personalization based on psychotype determination and the adaptation of VR content to the person's physiological parameters, which means that the personalization process is viewed as a two-level one.

Preliminary technical implementation of the approach showed its feasibility. Moreover, it has been revealed that some tension and monotony indexes decrease after correction. Based on this experience, further development of individualized VR correction of psychophysiological functioning is possible.

References:

1. World Health Organization. Mental health at work [Electronic resource]. - Geneva: WHO, 2024. - URL: https://www.who.int/news-room/fact-sheets/detail/mental-health-at-work (accessed: 29.04.2026).
2. Kim H., Kim D. J., Kim S., Chung W. H., Park K.-A., Kim J. D. K. et al. Effect of Virtual Reality on Stress Reduction and Change of Physiological Parameters Including Heart Rate Variability in People With High Stress: An Open Randomized Crossover Trial // Frontiers in Psychiatry. - 2021. - Vol. 12. - Article 614539. - DOI: 10.3389/fpsyt.2021.614539.
3. Spano G., Theodorou A., Reese G., Carrus G., Sanesi G., Panno A. Virtual nature, psychological and psychophysiological outcomes: A systematic review // Journal of Environmental Psychology. - 2023. - Vol. 89. - Article 102044. - DOI: 10.1016/j.jenvp.2023.102044.
4. Official website of ESKAL [Electronic resource]. - URL: https://эскал.рф (accessed: 29.04.2026).
5. Bilyi A. M., Belova D. A. Psychocorrection procedures based on personality types with virtual reality technologies // Vestnik psikhofiziologii. - 2020. - No. 1. - P. 90-93.
6. Vinnichenko A. D., Kotelnikova E. A., Redits A. V., Bilyi A. M. Methods for analysis and correction of human functional state with regard to temperament type // Vestnik psikhofiziologii. - 2021. - No. 3. - P. 67.
7. Belova D. A. The use of virtual reality technologies in psychocorrection based on personality types // ICCTI-2020 Proceedings. - 2020.
8. Great Lakes NeuroTechnologies. BioRadio User Guide [Electronic resource]. - URL: https://www.glneurotech.com/products/bioradio/ (accessed: 29.04.2026).
9. Schmidt P., Reiss A., Duerichen R., Marberger C., Van Laerhoven K. Introducing WESAD, a Multimodal Dataset for Wearable Stress and Affect Detection // Proceedings of the 20th ACM International Conference on Multimodal Interaction. - 2018. - P. 400-408. - DOI: 10.1145/3242969.3242985.
10. Zheng W.-L., Lu B.-L. A multimodal approach to estimating vigilance using EEG and forehead EOG // Journal of Neural Engineering. - 2017. - Vol. 14. - No. 2. - Article 026017. - DOI: 10.1088/1741-2552/aa5a98.
11. SEED Dataset. SEED-VIG dataset description [Electronic resource] / Shanghai Jiao Tong University, BCMI Laboratory. - URL: https://bcmi.sjtu.edu.cn/home/seed/seed-vig.html (accessed: 29.04.2026).
12. Wascher E., Heppner H., Hoffmann S. Evaluating Pro- and Re-Active Driving Behavior by Means of the EEG // Frontiers in Human Neuroscience. - 2018. - Vol. 12. - Article 205. - DOI: 10.3389/fnhum.2018.00205.