Definition of Emotional States Interval for Application of Artificial Intelligence and Stress Estimation

Definition of Emotional States Interval for Application of Artificial Intelligence and Stress Estimation Wilhelm Daniel Scherz1,2(B), Juan J. Perea2, Ralf Seepold1, and Juan Antonio Ortega3 1HTWG Konstanz, Alfred-Wachtel-Str. 8, 78462 Konstanz, Germany [email protected] 2Universidad de Loyola Andalucía, Av. de Las Universidades, 2, 41704 Dos Hermanas, Sevilla, Spain 3University of Seville, Avda. Reina Mercedes S/N, Seville, Spain Abstract. The perception of the amount of stress is subjective to every person, and the perception of it changes depending on many factors. One of the factors that has an impact on perceived stress is the emotional state. In this work, we compare the emotional state of 40 German driving students and present different partitions that can be advantageous for using artificial intelligence and classification. Like this, we evaluate the data quality and prepare for the specific use. The Stress Perceived Questionnaire (PSQ20) was employed to assess the level of stress experienced by individuals while participating in a driving simulation for 5 and 25 min. As a result of our analysis, we present a categorisation of various emotional states into intervals, comparing different classifications and facilitating a more straightforward implementation of artificial intelligence for classification purposes. Keywords: Stress ·Stress perceived questionnaire ·Driving 1 Introduction Stress has emerged as a topic of immense significance and interest in contemporary society [1], and it has been recognised by organisations like the WHO even as a disease [1]. Stress significantly influences driving behaviour, with a weighty influence on security and safety during driving and how people drive [2]. Drivers may feel stressed when they feel like they are not in control of their vehicle or the driving situation. Stress can be caused by various factors, such as traffic, bad weather, aggressive drivers or emotional state. Additional factors that also have an essential influence on rod safety are stress, fatigue and sleepiness [3, 4]. The German Federal Statistical Office (Destatis) reported 2,569 fatalities due to traffic accidents in Germany in 2021. This number decreased to 2,719 in 2020 but increased to 3,046 in 2019, and according to the World Health Organization, approximately 1.3 million people die each year due to traffic accidents worldwide [5]. The role of stress in https://doi.org/10.1007/978-3-031-48121-5_48 340 W. D. Scherz et al. risk evaluation is significant, making it an essential factor to understand. Studies show that stress can impair cognitive functions [6], it can lead to risky driving behaviour [7] and make it challenging to judge risks [8,9]. Stress can be defined as anything that disturbs or affects the internal equilibrium of the body, and a stressor is an external stimuli, perceived threads or demands that can create stress. This balance is called homeostasis. In addition to external stressors, there are also internal stressors. Internal stressors such as worries can also trigger the stress response of the body and disturb the internal equilibrium [10]. Among the widely known adverse effects of stress on the body are: degeneration of the cardiovascular system [11, 12], the inappropriate response of the immune system [13,14] and reduced capacity to cope with physical and mental demands [15]. Generally, stress is often categorised into the following characteristics: Insensitivity, duration, type and perception. This research aims to introduce a method for dividing perceived stress obtained from the PSQ20 into discrete intervals, thereby optimising the classification process across diverse sections enabling the use of advanced machine learning and artificial intelligence algorithms in the future for the classification of stress depending on the emotional state. The research should help to support a more straightforward implementation of artificial intelligence. 2 State of the Art In todays society, health awareness is increasingly important, and the collection of vital biological data is becoming more ordinary. This awareness is also true in the context of stress management, as people are seeking new ways to track and manage their stress levels [16]. There are mainly different methods for measuring stress. Those methods can be resumed in these categories. Questioners and self-report. This method asks the participants to rate their stress levels. Usually, it is done using questionnaires or a journal. Some examples of this include the Perceived Stress Questionnaire (PSQ20) [17] and the Perceived Stress Scale (PSS) [18]. Measurement by biochemical hormones: This method uses the advantage that stress releases hormones, i.e. cortisol and adrenalin. These hormones can be measured by taking blood, urine or saliva [19]. Sensorics measurements. Like the last method, stress also influences the behaviour of different physiological systems. Examples include heart rate variability, skin conductivity [20] and more physiological responses. Behavioural measurement. This method uses the subtle differences in behaviour between when a person is stressed or not is stressed. A good example of behavioural measurement is stress estimation by analysing the typing style [21]. Nonetheless, using a restricted set of parameters can pose challenges in distinguishing between healthy physical activity and adverse factors like stress. The PSQ20 questionnaire offers a distinct advantage in its simplicity when used for stress estimation and can be added to every data collection. Definition of Emotional States Interval for Application 341 3 Methodology To obtain the needed data to categorise various emotional states into intervals, a group of German students with driving experience was selected (n =40). To collect biovital data, a polar H10 chest strap was given to each participant. The focus was on the Heart Rate and the R-R Intervals (interbeat intervals). Before starting the experiment, the procedure was explained in detail to the participants and the voluntary agreement was collected. At the beginning of the experiment, the PSQ20 questionnaire was given to the participants, and they had to answer it. There was no time limit to answer the questionnaire. After the questionnaire was filled, the participant could sit on the simulator and listen to 5 min of relaxing music. After this, the participant had to start driving for 25 min. During this phase, different traffic challenges were used to act as stressors. A total of 35 male and 5 female students in Germany participated in this study. The mean age of the participants was 25.84 ±4.83. All participants had experience driving and a valid driving license. The participants were recruited voluntarily and had no financial compensation or benefit. The participants had to fulfil the following criteria: Subjects had to be healthy, have no history of heart problems, have good vision or use glasses to correct the vision. Additionally, the participants must be sober and did not consume drugs or medicaments before the experiment. 3.1 Driving Simulator The City Car Driving simulator was used for this experiment. The main advantages of this simulator are the advanced and realistic car physics that provides a reasonable degree of realism. Additionally, the traffic conditions can be configured. For this experiment, the traffic was set to normal, but events like car driving the opposite line, accidents, and randomly crossing pedestrians were activated. Additionally, when the driver is not driving according to the driving rules, a signal is released informing him about his mistake. 3.2 Perceived Stress Questionnaire The Perceived Stress Questionnaire (PSQ20) [17] is a frequently used questionnaire for estimating perceived stress. There are two forms of the questionnaire: the regular one with about 30 questions and the shorter one with 20 Questions. The shortened version of the questionnaire evaluates the following four emotional states: Worries, Tension, Joy and Demand. Higher scores in the PSQ20 mean that this emotional state is stronger. Dividing this in the quantiles helps change the score from a numerical value to intervals that are more suited for machine learning (ML) approaches. 342 W. D. Scherz et al. 4 Results The results intend to present a division of the perceived stress into different intervals to enable a more efficient classification into different regions using machine learning / artificial intelligence algorithms. Only complete questionnaires, participants with signed agreements and recorded biovital data were analysed. Because of these requirements, three datasets were discarded, reducing the sample to 40 participants. Figure 1divides the participants into intervals of 5 quantiles. Figure 2divides the participants into four quantiles, and Fig. 3into three quantiles. As we can observe, the differences between the worries quantiles are more significant than those in Figs. 2and 3. 0-20% 20.01- 40.00% 40.01- 60.00% 60.01- 80.00% 80.01- 100.00 % Worries 819102 1 overall score 221107 0 0 5 10 15 20 25 Fig. 1. Intervals of five quantiles A similar pattern is evident for the overall score in Figs. 1,2and 3. This suggests that the discernible differences between the various intervals may suffice for training artificial intelligence to perform classification tasks. Definition of Emotional States Interval for Application 343 0-25% 25.01- 50.00% 50.01- 75.00% 75.01- 100.00% Worries 14 16 8 2 overall score 7248 1 0 5 10 15 20 25 Fig. 2. Intervals of four quantiles 0-33.33 33.34-66.67 66.68-100 Worries 22 15 3 overall score 13 24 3 0 5 10 15 20 25 Fig. 3. Intervals of tree quantiles 5 Conclusion We have conducted a thorough study to quantify the stress levels of each patient in different domains, such as worries, demands, tension, joy, and overall state. Furthermore, we have developed a classification of the values in these domains for each patient and compared different divisions of these values: three, four, and five divisions. Our analysis 344 W. D. Scherz et al. has shown that the most relevant domains are worries and overall state. Based on the hypothesis that more significant variability provides more information, we have shown that the optimal classification is achieved with five divisions, as it maximises the information due to its higher variability. This behaviour and this division could be used for the age range of 25.84 ±4.83 of healthy persons. A validation with other ranges of ages could prove this separation as applicable. Looking at Fig. 1we can observe that more participants are located in the second quantile, telling us there is a bigger group of drivers (21 drivers) with a medium lower amount of stress. If we look at Fig. 3we can observe that most divers (24 drives) fit in the medium quantile, telling us that most drivers are in the middle. Comparing Figs. 1 and 3we can conclude that the smaller division in more quantiles can be advantageous for a more price description of the drivers. 6 Future Works Once we have established and justified this arrangement for classifying the stress levels of patients, a future line of research could involve using this classification to train artificial intelligence systems. The process would involve training the system with the proposed classification and, based on its variability, analysing data to classify patients. The final goal would be to develop a system capable of automatically classifying patients into one of the five categories, thereby determining their stress levels without requiring direct human intervention. This would restructure the analysis of patients stress levels and could be complemented with other analytical techniques, such as the Poincaré Plot. In addition to the collected biovital parameters and the information obtained from the questionnaires, additional information like demographic factors could deepen and help to understand deeper the nuances and characteristics of stress and enable a better classification. References 1. Andrade L, Caraveo-Anduaga JJ, Berglund P, Bijl R, Kessler RC, Demler O, Walters E, Kylyc C, Offord D, Ustun TB, Wittchen HU (2000) Cross-national comparisons of the prevalences and correlates of mental disorders. WHO Int Consort Psychiatr Epidemiol. Bull World Health Organiz 78:413–426 2. Yu Z, Yu T, Ge Y, Qu W (2023) The effect of perceived global stress and altruism on prosocial driving behavior, yielding behavior, and yielding attitude. Traffic Inj Prev 24:402–408. https:// doi.org/10.1080/15389588.2023.2191765 3. Radun I et al (2022) Sleepy drivers on a slippery road: a pilot study using a driving simulator. J Sleep Res 31:e13488. https://doi.org/10.1111/jsr.13488 4. Taylor AH, Dorn L (2006) Stress, fatigue, health, and risk of road traffic accidents among professional drivers: the contribution of physical inactivity. Annu Rev Public Health 27:371– 391. https://doi.org/10.1146/annurev.publhealth.27.021405.102117 5. Useche SA, Ortiz VG, Cendales BE (2017) Stress-related psychosocial factors at work, fatigue, and risky driving behavior in bus rapid transport (BRT) drivers. Accident Anal Prevent 104:106–114. https://doi.org/10.1016/j.aap.2017.04.023 Definition of Emotional States Interval for Application 345 6. Kulshreshtha A, Alonso A, McClure LA, Hajjar I, Manly JJ, Judd S (2023) Association of stress with cognitive function among older black and white US adults. JAMA Netw Open 6:e231860. https://doi.org/10.1001/jamanetworkopen.2023.1860 7. Delhomme P, Gheorghiu A (2021) Perceived stress, mental health, organisational factors, and self-reported risky driving behaviors among truck drivers circulating in France. J Safety Res 79:341–351. https://doi.org/10.1016/j.jsr.2021.10.001 8. Pooladvand S, Hasanzadeh S (2023) Impacts of stress on workers’ risk-taking behaviors: cognitive tunneling and impaired selective attention. J Constr Eng Manag 149. https://doi. org/10.1061/JCEMD4.COENG-13339 9. Reynolds EK, Schreiber WM, Geisel K, MacPherson L, Ernst M, Lejuez CW (2013) Influence of social stress on risk-taking behavior in adolescents. J Anxiety Disord 27:272–277. https:// doi.org/10.1016/j.janxdis.2013.02.010 10. Morrissey C et al (2023) Fear of cancer recurrence inventory scores and their correlation with quality of life and stress levels in breast cancer survivors. JCO 41:e24073–e24073. https:// doi.org/10.1200/JCO.2023.41.16_suppl.e24073 11. Zaffalon Júnior JR, Viana AO, Melo GEL, de Angelis K (2018) The impact of sedentarism on heart rate variability (HRV) at rest and in response to mental stress in young women. Physiol Rep 6:e13873. https://doi.org/10.14814/phy2.13873 12. Mei Y, Thompson MD, Cohen RA, Tong X (2015) Autophagy and oxidative stress in cardiovascular diseases. Biochem Biophys Acta 1852:243–251. https://doi.org/10.1016/j.bbadis. 2014.05.005 13. Martínez Fernández J, Augusto JC, Trombino G, Seepold R, Martinez Madrid N (2013) Self-aware trader: a new approach to safer trading. J Univ Comput Sci 14. Barrows IR, Ramezani A, Raj DS (2019) Inflammation, immunity, and oxidative stress in hypertension-partners in crime? Adv Chronic Kidney Dis 26:122–130. https://doi.org/10. 1053/j.ackd.2019.03.001 15. Cody K, Scott JM, Simmer-Beck M (2022) Examining the mental health of university students: a quantitative and qualitative approach to identifying prevalence, associations, stressors, and interventions. J Am College Health: J ACH 1–11. https://doi.org/10.1080/07448481.2022. 2057192 16. RaviPrakash H, Anwar SM (2023) Do it yourself. In: Byrne MF, Parsa N, Greenhill AT, Chahal D, Ahmad O, Bagci U (eds) AI in clinical medicine. Wiley, pp 94–103. https://doi. org/10.1002/9781119790686.ch10 17. Fliege H et al (2005) The perceived stress questionnaire (PSQ) reconsidered: validation and reference values from different clinical and healthy adult samples. Psychosom Med 67:78–88. https://doi.org/10.1097/01.psy.0000151491.80178.78 18. Lee E-H (2012) Review of the psychometric evidence of the perceived stress scale. Asian Nurs Res 6:121–127. https://doi.org/10.1016/j.anr.2012.08.004 19. El-Farhan N, Rees DA, Evans C (2017) Measuring cortisol in serum, urine and saliva—Are our assays good enough? Ann Clin Biochem 54:308–322. https://doi.org/10.1177/000456321 6687335 20. Najström M, Jansson B (2007) Skin conductance responses as predictor of emotional responses to stressful life events. Behav Res Ther 45:2456–2463. https://doi.org/10.1016/ j.brat.2007.03.001 21. Gunawardhane SDW, Silva PM, de Kulathunga DSB, Arunatileka SMKD (2013) Non invasive human stress detection using key stroke dynamics and pattern variations. In: 2013 international conference on advances in ICT for emerging regions (ICTer). IEEE, pp 240–247. https://doi. org/10.1109/ICTer.2013.6761185