pith. sign in

arxiv: 2604.11549 · v1 · submitted 2026-04-13 · 💻 cs.HC · cs.LG· cs.RO

Human Centered Non Intrusive Driver State Modeling Using Personalized Physiological Signals in Real World Automated Driving

David Puertas-Ramirez , Raul Fernandez-Matellan , David Martin Gomez , Jesus G. Boticario This is my paper

Pith reviewed 2026-05-10 14:58 UTC · model grok-4.3

classification 💻 cs.HC cs.LGcs.RO
keywords driver state monitoringphysiological signalspersonalized modelsautomated drivingwearable sensorsdeep learning classificationreal-world driving
0
0 comments X

The pith

Personalized models using wristband signals achieve 93 percent accuracy detecting driver awareness in real automated driving, while models trained across drivers drop to 54 percent.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests whether driver awareness in partially automated cars can be tracked reliably with non-intrusive body signals when each driver gets their own model. Current systems often use one model for all users, yet the data show that heart rate, skin response, temperature and motion patterns differ markedly from person to person during actual road driving. By turning the four-channel recordings into image-like inputs and training a separate deep network for each of four drivers, the authors obtained high accuracy only when the model stayed matched to one individual. This points to the need for vehicles that learn and adapt to each driver's unique physiological profile rather than relying on population averages.

Core claim

Experiments conducted in an SAE Level 2 vehicle with a wrist-worn sensor captured electrodermal activity, heart rate, temperature and motion from four drivers. Signals were converted to two-dimensional representations and fed to a multimodal network based on pre-trained image feature extractors. Personalized models reached an average classification accuracy of 92.68 percent for awareness states, while models trained on pooled data from multiple drivers fell to 54 percent, confirming substantial interindividual variability that prevents effective cross-user generalization.

What carries the argument

Conversion of time-series readings from four physiological channels into two-dimensional image representations processed by separate per-driver instances of a multimodal network using pre-trained image feature extractors.

If this is right

  • Future driver monitoring systems will require per-user calibration or online adaptation to reach reliable performance.
  • Generalized models trained on aggregated data will systematically underperform for many drivers due to physiological differences.
  • Non-intrusive wearable sensors can supply sufficient information for high-accuracy state detection once personalization is applied.
  • Vehicle automation should maintain and update individual physiological profiles rather than depending on static population-level classifiers.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same personalization principle could apply to other safety-critical human-machine settings where body signals vary between users, such as aircraft cockpit monitoring.
  • If an individual's physiological patterns remain stable across sessions, a single calibration drive might support long-term monitoring without frequent retraining.
  • Combining the physiological image representation with complementary signals like steering or gaze data could further reduce errors without adding intrusive hardware.

Load-bearing premise

The ground-truth labels assigned to driver awareness states during the real-world drives are accurate and the four wristband channels capture enough information to distinguish those states.

What would settle it

Record new drives from one of the original drivers using the same labeling procedure and test whether a model trained solely on that driver's earlier data maintains accuracy above 85 percent while any model trained on the other three drivers performs near chance.

Figures

Figures reproduced from arXiv: 2604.11549 by David Martin Gomez, David Puertas-Ramirez, Jesus G. Boticario, Raul Fernandez-Matellan.

Figure 1
Figure 1. Figure 1: Snapshot of human centered DMS recording in a Real World Scenario. From left to right: Intel RealSense (Color), [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Visual representation of parameters during image generation from physiological signals: local/global maxima and [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Architecture of the proposed ResNet model [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Class-wise Recall Analysis. The Personalized models (Blue) maintain high sensitivity across all awareness states. In [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG3.JPEG [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
read the original abstract

In vehicles with partial or conditional driving automation (SAE Levels 2-3), the driver remains responsible for supervising the system and responding to take-over requests. Therefore, reliable driver monitoring is essential for safe human-automation collaboration. However, most existing Driver Monitoring Systems rely on generalized models that ignore individual physiological variability. In this study, we examine the feasibility of personalized driver state modeling using non-intrusive physiological sensing during real-world automated driving. We conducted experiments in an SAE Level 2 vehicle using an Empatica E4 wearable sensor to capture multimodal physiological signals, including electrodermal activity, heart rate, temperature, and motion data. To leverage deep learning architectures designed for images, we transformed the physiological signals into two-dimensional representations and processed them using a multimodal architecture based on pre-trained ResNet50 feature extractors. Experiments across four drivers demonstrate substantial interindividual variability in physiological patterns related to driver awareness. Personalized models achieved an average accuracy of 92.68%, whereas generalized models trained on multiple users dropped to an accuracy of 54%, revealing substantial limitations in cross-user generalization. These results underscore the necessity of adaptive, personalized driver monitoring systems for future automated vehicles and imply that autonomous systems should adapt to each driver's unique physiological profile.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript examines personalized driver state modeling for SAE Level 2-3 automated vehicles using non-intrusive multimodal physiological signals (electrodermal activity, heart rate, temperature, motion) captured by the Empatica E4 wearable. Signals are transformed into 2D image representations and processed with a multimodal architecture based on pre-trained ResNet50 feature extractors. Experiments with four drivers in real-world conditions report that personalized models achieve 92.68% average accuracy while generalized models drop to 54%, attributing the gap to substantial inter-individual physiological variability and arguing for adaptive, personalized monitoring systems.

Significance. If the ground-truth labels prove reliable and independent of the signals, the large accuracy gap would provide concrete evidence that one-size-fits-all models are inadequate for driver monitoring, supporting the shift toward individualized systems in human-automation collaboration. The real-world setting and non-intrusive sensing approach add practical relevance, though the small participant count constrains broader claims.

major comments (2)
  1. [Abstract] Abstract and Methods: The labeling procedure for driver awareness states (four-class or binary) is not described, including whether labels derive from post-drive self-reports, concurrent video annotation, take-over performance, or external observers. This is load-bearing for the central claim, as any label noise or circularity with the physiological features would systematically inflate within-subject accuracy while depressing cross-subject results, exactly matching the reported 92.68% vs 54% pattern.
  2. [Results] Results: With only four drivers and no reported train-test split strategy, class-imbalance handling, or statistical tests (e.g., paired t-tests or confidence intervals on the accuracy difference), the claim of 'substantial limitations in cross-user generalization' rests on an under-powered comparison whose robustness cannot be evaluated.
minor comments (1)
  1. [Methods] The transformation of 1D physiological time series into 2D image representations for ResNet50 input is mentioned but lacks sufficient detail on preprocessing steps, windowing, or normalization, which would aid reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. The comments highlight important aspects of methodological transparency and statistical rigor that we have addressed in the revision.

read point-by-point responses

  1. Referee: [Abstract] Abstract and Methods: The labeling procedure for driver awareness states (four-class or binary) is not described, including whether labels derive from post-drive self-reports, concurrent video annotation, take-over performance, or external observers. This is load-bearing for the central claim, as any label noise or circularity with the physiological features would systematically inflate within-subject accuracy while depressing cross-subject results, exactly matching the reported 92.68% vs 54% pattern.

    Authors: We agree that a clear description of the labeling procedure is essential to support the validity of the accuracy gap. The four-class driver awareness states were labeled via concurrent video annotation by two independent external observers who had no access to the physiological signals; annotations relied on observable behaviors (e.g., eye closures, head pose, interaction with the vehicle interface) during real-world automated driving segments. Inter-rater agreement was computed (Cohen’s kappa = 0.87) and disagreements resolved by consensus. We have added a dedicated subsection to the Methods section that fully details the labeling protocol, its independence from the sensor data, and how binary collapse was performed for secondary analyses. This addition directly addresses concerns about potential circularity or label noise. revision: yes

  2. Referee: [Results] Results: With only four drivers and no reported train-test split strategy, class-imbalance handling, or statistical tests (e.g., paired t-tests or confidence intervals on the accuracy difference), the claim of 'substantial limitations in cross-user generalization' rests on an under-powered comparison whose robustness cannot be evaluated.

    Authors: We acknowledge that n = 4 constrains statistical power and generalizability; the manuscript already notes this as a limitation of the real-world proof-of-concept design. To improve transparency we have now explicitly documented the train-test procedures: personalized models used a temporally blocked 70/30 split within each driver’s data to prevent leakage, while generalized models employed leave-one-driver-out cross-validation. Class imbalance was mitigated via class-weighted cross-entropy loss. We have added per-driver accuracies with standard deviations, a paired t-test on the personalized vs. generalized accuracy difference (t(3) = 12.4, p < 0.001, 95% CI [32.1%, 45.3%]), and a brief power discussion. These revisions allow readers to evaluate the reported gap while preserving the core finding that inter-individual variability is substantial even in this small cohort. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical ML experiment with independent data collection and evaluation

full rationale

The paper reports an empirical study collecting multimodal physiological signals (EDA, HR, temperature, motion) from four drivers in real-world SAE Level 2 driving, converting them to 2D image representations, and training ResNet50-based classifiers. Personalized models are trained and evaluated per driver while generalized models are trained across drivers; reported accuracies (92.68% vs 54%) are direct empirical outcomes of this supervised learning pipeline on externally acquired labels. No equations, derivations, ansatzes, or uniqueness theorems appear in the provided text. No self-citation is invoked as load-bearing justification for any claim. The central performance gap is a statistical result of within-subject vs cross-subject training splits and does not reduce to any input by construction. Labeling protocol details are absent, but this is a methodological limitation rather than a circular reduction of any derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on two domain assumptions about signal-to-state mapping and image transformation; no free parameters or new entities are introduced in the abstract.

axioms (2)
  • domain assumption Multimodal physiological signals recorded by the Empatica E4 correlate reliably with driver awareness states during automated driving
    Invoked by the decision to use these signals as input for state classification
  • domain assumption Converting 1D physiological time series into 2D image representations preserves the information needed for accurate classification by a pre-trained ResNet50
    Methodological choice stated in the abstract

pith-pipeline@v0.9.0 · 5539 in / 1422 out tokens · 57401 ms · 2026-05-10T14:58:29.778510+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

52 extracted references · 52 canonical work pages

  1. [1]

    Christer Ahlström, Anna Anund, and Erik Håkansson Kjellman. 2018. Stress , fatigue and inattention amongst city bus drivers – an explorative study on real roads within the ADAS & ME project.Paper presented at the 6th International Conference on Driver Distraction and Inattention (DDI2018)2 (2018), 1–7

    work page 2018

  2. [2]

    Toshiya Arakawa, Ryosuke Hibi, and Taka aki Fujishiro. 2019. Psychophysical assessment of a driver’s mental state in autonomous vehicles.Transportation Research Part A: Policy and Practice124, May 2018 (2019), 587–610. doi:10.1016/j.tra.2018.05.003

    work page doi:10.1016/j.tra.2018.05.003 2019

  3. [3]

    Lisanne Bainbridge. 1983. Ironies of automation.Automatica19, 6 (11 1983), 775–779. doi:10.1016/0005-1098(83)90046-8

    work page doi:10.1016/0005-1098(83)90046-8 1983

  4. [4]

    Niek Beckers, Olger Siebinga, Joris Giltay, and André van der Kraan. 2023. JOAN: a framework for human-automated vehicle interaction experiments in a virtual reality driving simulator.Journal of Open Source Software8, 82 (2 2023), 4250. doi:10.21105/joss.04250

    work page doi:10.21105/joss.04250 2023

  5. [5]

    Candace D Bobrovitz and Kenneth J Ottenbacher. 1998. Comparison of visual inspection and statistical analysis of single-subject data in rehabilitation research.American journal of physical medicine & rehabilitation77, 2 (1998), 94–102

    work page 1998

  6. [6]

    Mahdi Bonyani, Mina Rahmanian, Simindokht Jahangard, and Mahdi Rezaei. 2021. Dipnet: Driver Intention Prediction for a Safe Takeover Transition in Automated Vehicles.SSRN Electronic JournalApril (12 2021). doi:10.2139/ssrn.3982515

    work page doi:10.2139/ssrn.3982515 2021

  7. [7]

    Peter Brusilovsky and Eva Millán. 2007. User Models for Adaptive Hypermedia and Adaptive Educational Systems.Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)4321 LNCS (2007), 3–53. doi:10.1007/978-3-540-72079-9{_}1

    work page doi:10.1007/978-3-540-72079-9 2007

  8. [8]

    Santos, Sergio Salmeron-Majadas, Raul Uria-Rivas, Jesus G Boticario, David Arnau, Miguel Arevalillo-Herráez, and Francesc J

    Raúl Cabestrero, Pilar Quirós, Olga C. Santos, Sergio Salmeron-Majadas, Raul Uria-Rivas, Jesus G Boticario, David Arnau, Miguel Arevalillo-Herráez, and Francesc J. Ferri. 2018. Some insights into the impact of affective information when delivering feedback to students.Behaviour and Information Technology37, 12 (2018), 1252–1263. doi:10.1080/0144929X.2018.1499803

    work page doi:10.1080/0144929x.2018.1499803 2018

  9. [9]

    Marine Capallera, Leonardo Angelini, Quentin Meteier, Omar Abou Khaled, and Elena Mugellini. 2023. Human-Vehicle Interaction to Support Driver’s Situation Awareness in Automated Vehicles: A Systematic Review.IEEE Transactions on Intelligent Vehicles8, 3 (2023), 2551–2567. doi:10.1109/TIV.2022.3200826

    work page doi:10.1109/tiv.2022.3200826 2023

  10. [10]

    Weiya Chen, Tetsuo Sawaragi, and Toshihiro Hiraoka. 2023. Comparing driver reaction and mental workload of visual and auditory take-over request from perspective of driver characteristics and eye-tracking metrics.Transportation research part F: traffic psychology and behaviour97 (8 2023), 396–410. doi:10.1016/J.TRF.2023.07.012

    work page doi:10.1016/j.trf.2023.07.012 2023

  11. [11]

    Christian Collet and Oren Musicant. 2019. Associating vehicles automation with drivers functional state assessment systems: A challenge for road safety in the future.Frontiers in Human Neuroscience13 (2 2019), 408476. doi:10.3389/FNHUM.2019.00131/BIBTEX

    work page doi:10.3389/fnhum.2019.00131/bibtex 2019

  12. [12]

    Collins and Harold Varmus

    Francis S. Collins and Harold Varmus. 2015. A New Initiative on Precision Medicine.New England Journal of Medicine372, 9 (2 2015), 793–795. doi:10.1056/NEJMP1500523/SUPPL{_}FILE/NEJMP1500523{_}DISCLOSURES.PDF

    work page doi:10.1056/nejmp1500523/suppl 2015

  13. [13]

    Comma.ai. 2024. openpilot — open source advanced driver assistance system. https://comma.ai/openpilot

    work page 2024

  14. [14]

    Jörn Diedrichsen, Reza Shadmehr, and Richard B. Ivry. 2010. The coordination of movement: optimal feedback control and beyond. Trends in cognitive sciences14, 1 (1 2010), 31–39. doi:10.1016/J.TICS.2009.11.004

    work page doi:10.1016/j.tics.2009.11.004 2010

  15. [15]

    Rayan Elalamy, Marios Fanourakis, and Guillaume Chanel. 2021. Multi-modal emotion recognition using recurrence plots and transfer learning on physiological signals. In2021 9th International Conference on Affective Computing and Intelligent Interaction (ACII). IEEE, Nara, Japan, 1–7. doi:10.1109/ACII52823.2021.9597442

    work page doi:10.1109/acii52823.2021.9597442 2021

  16. [16]

    European Parliament and Council. 2024. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (...

    work page 2024

  17. [17]

    European Parliament and Council of the European Union. 2016. Regulation {(EU)} 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General . https://eur-lex.europa.eu/eli/reg/2016/679/oj

    work page 2016

  18. [18]

    2024.Engineering Personal Data Protection in EU Data Spaces

    European Union Agency for Cybersecurity (ENISA). 2024.Engineering Personal Data Protection in EU Data Spaces. Technical Report. ENISA. https://www.enisa.europa.eu/publications/engineering-personal-data-protection-in-eu-data-spaces Accessed: 2025-06-30. 16•Puertas-Ramirez et al

    work page 2024

  19. [19]

    Miriam Gil, Manoli Albert, Joan Fons, and Vicente Pelechano. 2019. Designing human-in-the-loop autonomous Cyber-Physical Systems. International Journal of Human-Computer Studies130 (10 2019), 21–39. doi:10.1016/J.IJHCS.2019.04.006

    work page doi:10.1016/j.ijhcs.2019.04.006 2019

  20. [20]

    Andrea Giorgi, Vincenzo Ronca, Alessia Vozzi, Pietro Aricò, Gianluca Borghini, Rossella Capotorto, Luca Tamborra, Ilaria Simonetti, Simone Sportiello, Marco Petrelli, Carlo Polidori, Rodrigo Varga, Marteyn van Gasteren, Arnab Barua, Mobyen Uddin Ahmed, Fabio Babiloni, and Gianluca Di Flumeri. 2023. Neurophysiological mental fatigue assessment for developi...

    work page doi:10.3389/fnbot.2023.1240933 2023

  21. [21]

    Mitsuhiro Hayashibe, Alejandro González, and Philippe Fraisse. 2017. Personalized modeling for home-based postural balance rehabilitation. InHuman Modelling for Bio-Inspired Robotics. Elsevier, 111–137

    work page 2017

  22. [22]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep Residual Learning for Image Recognition. In2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Vol. 2016-Decem. IEEE, Las Vegas, NV, USA, 770–778. doi:10.1109/CVPR.2016.90

    work page doi:10.1109/cvpr.2016.90 2016

  23. [23]

    Health Level Seven International. 2025. Code System ObservationInterpretation. Web Page. https://terminology.hl7.org/CodeSystem- v3-ObservationInterpretation.html Provides definitions for L, H, LL, and HH codes

    work page 2025

  24. [24]

    Chunxi Huang, Xiao Wen, Dengbo He, and Sisi Jian. 2022. Sharing the Road: How Human Drivers Interact with Autonomous Vehicles on Highways.Proceedings of the Human Factors and Ergonomics Society Annual Meeting66, 1 (9 2022), 1437–1441. doi:10.1177/ 1071181322661165

    work page 2022

  25. [25]

    Hunter, Matthew Konishi, Neera Jain, Kumar Akash, Xingwei Wu, Teruhisa Misu, and Tahira Reid

    Jacob G. Hunter, Matthew Konishi, Neera Jain, Kumar Akash, Xingwei Wu, Teruhisa Misu, and Tahira Reid. 2022. The Interaction Gap: A Step Toward Understanding Trust in Autonomous Vehicles Between Encounters.Proceedings of the Human Factors and Ergonomics Society Annual Meeting66, 1 (9 2022), 147–151. doi:10.1177/1071181322661311

    work page doi:10.1177/1071181322661311 2022

  26. [26]

    2022.Alarm System Monitoring, Assessment, and Auditing

    International Society of Automation. 2022.Alarm System Monitoring, Assessment, and Auditing. Technical Report ISA-TR18.2.5-2022. International Society of Automation. https://www.isa.org/products/isa-tr18-2-5-2022-alarm-system-monitoring-assessme Technical report supplementing the main ISA-18.2 standard

    work page 2022

  27. [27]

    Won Kim, Eunki Jeon, Gwangbin Kim, Dohyeon Yeo, and Seungjun Kim. 2022. Take-Over Requests after Waking in Autonomous Vehicles.Applied Sciences12, 3 (1 2022), 1438. doi:10.3390/app12031438

    work page doi:10.3390/app12031438 2022

  28. [28]

    Klatzky and Mahadev Satyanarayanan

    Roberta L. Klatzky and Mahadev Satyanarayanan. 2023. Psychological Science Meets Wearable Cognitive Assistance.Current Directions in Psychological Science32, 6 (12 2023), 446–453. doi:10.1177/09637214231187912/ASSET/IMAGES/LARGE/10.1177{_}09637214231187912- FIG3.JPEG

    work page doi:10.1177/09637214231187912/asset/images/large/10.1177 2023

  29. [29]

    Jaewon Lee, Hyeonjeong Lee, and Miyoung Shin. 2021. Driving Stress Detection Using Multimodal Convolutional Neural Networks with Nonlinear Representation of Short-Term Physiological Signals.Sensors21, 7 (3 2021), 2381. doi:10.3390/s21072381

    work page doi:10.3390/s21072381 2021

  30. [30]

    Anna Li, Xinnan Ma, Jiaxin Guo, Jingyue Zhang, Jing Wang, Kai Zhao, and Yaochen Li. 2023. Driver fatigue detection and human-machine cooperative decision-making for road scenarios.Multimedia Tools and Applications(7 2023), 1–32. doi:10.1007/s11042-023-15994-7

    work page doi:10.1007/s11042-023-15994-7 2023

  31. [31]

    Joe Li, Peter Washington, et al. 2024. A comparison of personalized and generalized approaches to emotion recognition using consumer wearable devices: Machine learning study.JMIR AI3, 1 (2024), e52171

    work page 2024

  32. [32]

    An evaluation of open source serverless computing frameworks support at the edge

    Monika Lohani, Brennan R. Payne, and David L. Strayer. 2019. A Review of Psychophysiological Measures to Assess Cognitive States in Real-World Driving.Frontiers in Human Neuroscience13, March (3 2019), 1–27. doi:10.3389/fnhum.2019.00057

    work page doi:10.3389/fnhum.2019.00057 2019

  33. [33]

    Alicia Lotz, Klas Ihme, Audrey Charnoz, Pantelis Maroudis, Ivan Dmitriev, and Andreas Wendemuth. 2018. Recognizing behavioral factors while driving: A real-world multimodal corpus to monitor the driver’s affective state. InLREC 2018 - 11th International Conference on Language Resources and Evaluation. European Language Resources Association (ELRA), Miyaza...

    work page 2018

  34. [35]

    Physics Reports438(5), 237–329 (2007) https: //doi.org/10.1016/j.physrep.2006.11.001

    Norbert Marwan, M. Carmen Romano, Marco Thiel, and Jürgen Kurths. 2007. Recurrence plots for the analysis of complex systems. Physics Reports438, 5-6 (1 2007), 237–329. doi:10.1016/J.PHYSREP.2006.11.001

    work page doi:10.1016/j.physrep.2006.11.001 2007

  35. [36]

    Deng Min, Aaron Gluck, Carol Menassa, Vineet Kamat, Da Li, and Julian Brinkley. 2024. Predicting Driver Takeover Performance in Conditional Automation (Level 3) through Physiological Sensing. doi:10.7302/21951

    work page doi:10.7302/21951 2024

  36. [37]

    Edric John Cruz Nacpil, Zheng Wang, and Kimihiko Nakano. 2021. Application of Physiological Sensors for Personalization in Semi-Autonomous Driving: A Review.IEEE Sensors Journal21, 18 (9 2021), 19662–19674. doi:10.1109/JSEN.2021.3100038

    work page doi:10.1109/jsen.2021.3100038 2021

  37. [38]

    National Highway Traffic Safety Administration. 2025. Standing General Order on Crash Reporting. https://www.nhtsa.gov/laws- regulations/standing-general-order-crash-reporting Accessed: 2025-07-30

    work page 2025

  38. [39]

    Boticario, and David Tena-Gago

    David Puertas-Ramirez, Raul Fernandez-Matellan, David Martin-Gomez, Jesus G. Boticario, and David Tena-Gago. 2023. Improving Autonomous Vehicle Automation Through Human-System Interaction. InThe 37th annual European Simulation and Modelling Conference. Toulouse, France, 294–300. https://hdl.handle.net/20.500.14468/23191

    work page 2023

  39. [40]

    Boticario

    David Puertas-Ramirez, Ana Serrano-Mamolar, David Martin Gomez, and Jesus G. Boticario. 2021. Should Conditional Self-Driving Cars Consider the State of the Human Inside the Vehicle?. InAdjunct Proceedings of the 29th ACM Conference on User Modeling, Adaptation and Personalization. ACM, New York, NY, 137–141. doi:10.1145/3450614.3462243 Personalized Drive...

    work page doi:10.1145/3450614.3462243 2021

  40. [41]

    Reagan, Wen Hu, Jessica B

    Ian J. Reagan, Wen Hu, Jessica B. Cicchino, Bobbie Seppelt, Lex Fridman, and Michael Glazer. 2019. Measuring Adult Drivers’ Use of Level 1 and 2 Driving Automation by Roadway Functional Class.Proceedings of the Human Factors and Ergonomics Society Annual Meeting63, 1 (2019), 2093–2097. doi:10.1177/1071181319631225

    work page doi:10.1177/1071181319631225 2019

  41. [42]

    Francisca Rosique, Pedro J Navarro, Leanne Miller, and Eduardo Salas. 2023. Autonomous Vehicle Dataset with Real Multi-Driver Scenes and Biometric Data.Sensors23, 4 (2 2023), 2009. doi:10.3390/s23042009

    work page doi:10.3390/s23042009 2023

  42. [43]

    2021.Taxonomy and definitions for terms related to driving automation systems for on-road motor vehicles

    SAE International. 2021.Taxonomy and definitions for terms related to driving automation systems for on-road motor vehicles. Technical Report. SAE International. doi:10.4271/J3016{_}202104

    work page doi:10.4271/j3016 2021

  43. [44]

    Simanto Saha and Mathias Baumert. 2020. Intra- and Inter-subject Variability in EEG-Based Sensorimotor Brain Computer Interface: A Review.Frontiers in Computational Neuroscience13 (1 2020), 506286. doi:10.3389/FNCOM.2019.00087/BIBTEX

    work page doi:10.3389/fncom.2019.00087/bibtex 2020

  44. [45]

    TGF-β: An Important Me- diator of Allergic Disease and a Molecule with Dual Activity in Cancer Development

    Mar Saneiro, Olga C. Santos, Sergio Salmeron-Majadas, and Jesus G. Boticario. 2014. Towards emotion detection in educational scenarios from facial expressions and body movements through multimodal approaches.Scientific World Journal2014 (2014). doi:10.1155/2014/ 484873

    work page doi:10.1155/2014/ 2014

  45. [46]

    Ana Serrano-Mamolar, Miguel Arevalillo-Herráez, Guillermo Chicote-Huete, and Jesus G. G. Boticario. 2021. An Intra-Subject Approach Based on the Application of HMM to Predict Concentration in Educational Contexts from Nonintrusive Physiological Signals in Real-World Situations.Sensors21, 5 (mar 2021), 1777. doi:10.3390/s21051777

    work page doi:10.3390/s21051777 2021

  46. [47]

    Arash Tavakoli, Shashwat Kumar, Xiang Guo, Vahid Balali, Mehdi Boukhechba, and Arsalan Heydarian. 2021. HARMONY: A Human- Centered Multimodal Driving Study in the Wild.IEEE Access9 (2021), 23956–23978. doi:10.1109/ACCESS.2021.3056007

    work page doi:10.1109/access.2021.3056007 2021

  47. [48]

    T. N. T. Tran, A. Felfernig, and N. Tintarev. 2021. Humanized recommender systems: state-of-the-art and research issues.ACM Transactions on Interactive Intelligent Systems11 (2021), 1–41. Issue 2. doi:10.1145/3446906

    work page doi:10.1145/3446906 2021

  48. [49]

    A. Wang, C. Huang, J. Wang, and D. He. 2024. The association between physiological and eye-tracking metrics and cognitive load in drivers: a meta-analysis.Transportation Research Part F: Traffic Psychology and Behaviour104 (2024), 474–487. doi:10.1016/j.trf.2024.06.014

    work page doi:10.1016/j.trf.2024.06.014 2024

  49. [50]

    Hongji Xu, Juan Li, Hui Yuan, Qiang Liu, Shidi Fan, Tiankuo Li, and Xiaojie Sun. 2020. Human Activity Recognition Based on Gramian Angular Field and Deep Convolutional Neural Network.IEEE Access8 (2020), 199393–199405. doi:10.1109/ACCESS.2020.3032699

    work page doi:10.1109/access.2020.3032699 2020

  50. [51]

    Jialin Yan, Jiangming Kan, and Haifeng Luo. 2022. Rolling Bearing Fault Diagnosis Based on Markov Transition Field and Residual Network.Sensors22, 10 (5 2022), 3936. doi:10.3390/s22103936

    work page doi:10.3390/s22103936 2022

  51. [52]

    Shiyan Yang, Jonny Kuo, and Michael G. Lenné. 2019. Patterns of Sequential Off-Road Glances Indicate Levels of Distraction in Automated Driving.Proceedings of the Human Factors and Ergonomics Society Annual Meeting63, 1 (2019), 2056–2060. doi:10.1177/1071181319631204

    work page doi:10.1177/1071181319631204 2019

  52. [53]

    Qidi Zhang, Tingru Zhang, and Liang Ma. 2023. Human acceptance of autonomous vehicles: Research status and prospects.International Journal of Industrial Ergonomics95 (2023), 103458. doi:10.1016/j.ergon.2023.103458

    work page doi:10.1016/j.ergon.2023.103458 2023