pith. sign in

arxiv: 2604.04832 · v1 · submitted 2026-04-06 · 💻 cs.HC

When One Sensor Fails: Tolerating Dysfunction in Multi-Sensor Prototypes

Freek Hens , Amirhossein Sadough , Aleksa Bok\v{s}an , Mahyar Shahsavari , Mohammad Mahdi Dehshibi This is my paper

Pith reviewed 2026-05-10 20:21 UTC · model grok-4.3

classification 💻 cs.HC
keywords sEMGsensor failureFisher discriminant ratiogesture recognitionmulti-sensor systemsfail-safe designhuman-computer interactionsensor ablation
0
0 comments X

The pith

Systematic sensor ablations paired with Fisher discriminant ratio analysis produce a ranking of crucial versus replaceable sensors in sEMG arrays.

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

The paper proposes a framework to make multi-sensor surface electromyography systems tolerant to single-sensor failure. By removing sensors one at a time from arm recordings of rock-paper-scissors gestures, extracting hand-crafted features, and measuring how well the remaining sensors still separate the gesture classes, the authors obtain a clear ordering of sensor importance. A multi-layer perceptron confirms that the separability measures align with actual classification performance. This ordering directly tells designers which sensors must be kept or duplicated and which can be spared, supporting more reliable hardware for human-computer interaction and clinical use.

Core claim

Using arm sEMG recordings of rock-paper-scissors gestures, hand-crafted features were extracted and class separability quantified via the maximum Fisher discriminant ratio after systematic sensor ablations. The resulting ranking of crucial versus replaceable sensors, validated by a multi-layer perceptron and consistent with physiological evidence, provides a concrete basis for implementing fail-safe mechanisms, sensor redundancy, and robust device design in multi-sensor sEMG systems.

What carries the argument

Systematic one-by-one sensor ablation combined with maximum Fisher discriminant ratio (FDR) analysis to rank each sensor's contribution to gesture-class separability.

If this is right

  • Designers can identify which sensors must be retained or duplicated to preserve system usability after one failure.
  • Sensors ranked as replaceable can be removed or used only for redundancy, reducing hardware cost while keeping performance.
  • The same ablation and FDR procedure supplies a repeatable method for building fail-safe mechanisms into new multi-sensor prototypes.
  • The approach supports improved reliability in both clinical rehabilitation devices and everyday human-computer interaction applications.

Where Pith is reading between the lines

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

  • The FDR-based ranking could be recomputed on continuous or compound gestures to check whether different sensors become critical under sustained use.
  • Wearable-device makers could embed the ranking as a static map to trigger automatic signal substitution when a low-ranked sensor drops out.
  • Clinical teams might run per-user ablations to personalize which sensors receive backup wiring or higher sampling rates.

Load-bearing premise

The importance ranking derived from rock-paper-scissors gestures and the chosen hand-crafted features will generalize to other gestures, users, and real-world conditions without major loss of separability.

What would settle it

Running the identical ablation-plus-FDR procedure on a new set of gestures or a different participant cohort and finding that the top-ranked sensors change order.

Figures

Figures reproduced from arXiv: 2604.04832 by Aleksa Bok\v{s}an, Amirhossein Sadough, Freek Hens, Mahyar Shahsavari, Mohammad Mahdi Dehshibi.

Figure 1
Figure 1. Figure 1: : Fall back mechanism for sEMG-based gesture recognition. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: : (a) Conceptual illustration of class separability. The figure visualises the vary [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: : Conceptual illustration of the data collection setup using the Myo armband [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: : Stage 1 Results: (a) Our framework’s model-free analysis predicts that the [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: : Stage 2 Results: Per-class sensor criticality. A larger bar indicates a more [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
read the original abstract

Surface electromyography (sEMG) sensors are widely used in human-computer interaction, yet the failure of a single sensor can compromise system usability. We propose a methodological framework for implementing a fail-safe mechanism in multi-sensor sEMG systems. Using arm sEMG recordings of rock-paper-scissors gestures, we extracted hand-crafted features and quantified class separability via the maximum Fisher discriminant ratio (FDR). A multi-layer perceptron validated our approach, consistent with prior findings and physiological evidence. Systematic sensor ablations and FDR analysis produced a ranking of crucial versus replaceable sensors. This ranking informs robust device design, sensor redundancy, and reliability in clinical and practical applications.

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 proposes a methodological framework for fail-safe mechanisms in multi-sensor sEMG systems for human-computer interaction. Using arm sEMG recordings of rock-paper-scissors gestures, the authors extract hand-crafted features, quantify class separability with the maximum Fisher discriminant ratio (FDR), perform systematic sensor ablations to produce a ranking of crucial versus replaceable sensors, and validate the approach with a multi-layer perceptron (MLP). The resulting ranking is claimed to inform robust device design, sensor redundancy, and reliability in clinical and practical applications, consistent with prior findings and physiological evidence.

Significance. If the sensor importance ranking holds under broader conditions, the work could contribute to more reliable sEMG-based interfaces by identifying essential sensors for redundancy planning. The use of standard FDR metrics and ablation analysis provides a transparent, reproducible way to assess sensor contributions, which is a strength for practical HCI and clinical device design.

major comments (2)
  1. [Abstract] Abstract: The description of the approach is high-level and supplies no quantitative results, error bars, statistical tests, or details on feature definitions and MLP architecture. Without these, it is impossible to verify whether the FDR ranking actually supports the fail-safe claim.
  2. [Results] The central claim that the ranking informs robust device design assumes the relative sensor importance is robust to changes in gesture vocabulary, subject population, electrode placement, and feature choice. No cross-gesture validation, multi-subject leave-one-out tests, or comparisons against learned features or alternative separability metrics are described, so the ranking could be an artifact of the three-class RPS task and specific feature set.
minor comments (1)
  1. [Abstract] The abstract would benefit from specifying the number of sensors, subjects, and trials to allow readers to assess the scale of the experiments.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their constructive comments on our manuscript. We address each major comment point by point below, indicating where revisions will be made.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The description of the approach is high-level and supplies no quantitative results, error bars, statistical tests, or details on feature definitions and MLP architecture. Without these, it is impossible to verify whether the FDR ranking actually supports the fail-safe claim.

    Authors: We agree that the abstract is high-level and omits quantitative details. In the revised version, we will expand the abstract to report key FDR values for the sensor ranking (with standard deviations across runs), note any statistical tests performed, and briefly specify the hand-crafted features (e.g., time-domain statistics) and MLP architecture (number of layers and units). These additions will allow readers to assess the support for the fail-safe mechanism without altering the manuscript's core contribution. revision: yes

  2. Referee: [Results] The central claim that the ranking informs robust device design assumes the relative sensor importance is robust to changes in gesture vocabulary, subject population, electrode placement, and feature choice. No cross-gesture validation, multi-subject leave-one-out tests, or comparisons against learned features or alternative separability metrics are described, so the ranking could be an artifact of the three-class RPS task and specific feature set.

    Authors: The referee correctly identifies that our validation is confined to the rock-paper-scissors gesture set, the chosen hand-crafted features, and the available recordings. The manuscript does not include cross-gesture testing, leave-one-out multi-subject validation, or direct comparisons to learned features or other metrics such as mutual information. While the FDR ranking aligns with physiological evidence of muscle activation patterns for these gestures, we do not claim invariance across all conditions. In revision, we will add a limitations subsection explicitly discussing these scope constraints, include a comparison of FDR against an alternative separability measure, and clarify that the framework is presented as a generalizable method demonstrated on this task. New multi-subject data collection or expanded gesture experiments cannot be performed in this revision. revision: partial

standing simulated objections not resolved
  • New multi-subject data collection and cross-gesture validation experiments, which would require additional recordings beyond the current study.

Circularity Check

0 steps flagged

No circularity: empirical ablation ranking is data-driven

full rationale

The paper's core procedure—systematic sensor ablations followed by FDR computation on hand-crafted sEMG features from rock-paper-scissors data—directly measures class separability after each removal. FDR is an externally defined, non-circular separability metric; the resulting ranking is an output of that computation rather than a redefinition or fit of the input data. No equations, self-citations, or ansatzes are shown that would force the ranking to equal its own inputs by construction. The MLP validation step is independent and does not close any loop back to the FDR ranking.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides insufficient detail to enumerate specific free parameters, axioms, or invented entities. The approach appears to rest on standard signal-processing assumptions and the validity of FDR as a separability measure, none of which are derived or justified within the given text.

pith-pipeline@v0.9.0 · 5426 in / 1215 out tokens · 35758 ms · 2026-05-10T20:21:13.438960+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

  1. [1]

    Digital Twin

    Tanish Baranwal, Srihari Varada, Santanu Das, and Mohammad R. Haider. Fault- Tolerant IoT System Using Software-Based “Digital Twin”. In 2024 IEEE 10th World Forum on Internet of Things (WF-IoT) , pages 828–833. IEEE, 2024

    work page 2024

  2. [2]

    A memristive associative learning circuit for fault-tolerant multi-sensor fusion in autonomous vehicles

    Kapil Bhardwaj, Dmitrii Semenov, Roman Sotner, and Sayani Majumdar. A memristive associative learning circuit for fault-tolerant multi-sensor fusion in autonomous vehicles. Advanced Intelligent Systems, page 2500215, 2025

    work page 2025

  3. [3]

    McDonald, Tammy Toscos, Mike Y

    Sunny Consolvo, David W. McDonald, Tammy Toscos, Mike Y. Chen, Jon Froehlich, Beverly Harrison, Predrag Klasnja, Anthony LaMarca, Louis LeGrand, Ryan Libby, Ian Smith, and James A. Landay. Activity sensing in the wild: a field trial of ubifit garden. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems , pages 1797–1806. Associati...

    work page 2008

  4. [4]

    ADVISE: ADaptive feature relevance and VISual Explanations for convolutional neural networks

    Mohammad Mahdi Dehshibi, Mona Ashtari-Majlan, Gereziher Adhane, and David Masip. ADVISE: ADaptive feature relevance and VISual Explanations for convolutional neural networks. The Visual Computer , 40(8):5407–5419, 2024

    work page 2024

  5. [5]

    Pain level and pain-related behaviour classification using gru-based sparsely-connected rnns

    Mohammad Mahdi Dehshibi, Temitayo Olugbade, Fernando Diaz-de Maria, Nadia Bianchi-Berthouze, and Ana Tajadura-Jim´ enez. Pain level and pain-related behaviour classification using gru-based sparsely-connected rnns. IEEE Journal of Selected Topics in Signal Processing, 17(3):677–688, 2023

    work page 2023

  6. [6]

    Emg from forearm datasets for hand gestures recognition, May 2019

    Elisa Donati. Emg from forearm datasets for hand gestures recognition, May 2019

    work page 2019

  7. [7]

    Ronald A. Fisher. The use of multiple measurements in taxonomic problems. Annals of Eugenics, 7(2):179–188, 1936

    work page 1936

  8. [8]

    A myoelectric prosthetic hand with muscle synergy–based motion determination and impedance model–based biomimetic control

    Akira Furui, Shintaro Eto, Kosuke Nakagaki, Kyohei Shimada, Go Nakamura, Akito Masuda, Takaaki Chin, and Toshio Tsuji. A myoelectric prosthetic hand with muscle synergy–based motion determination and impedance model–based biomimetic control. Science Robotics, 4(31):eaaw6339, 2019. 12

    work page 2019

  9. [9]

    Signals to spikes for neuromorphic regulated reservoir computing and emg hand gesture recognition

    Nikhil Garg, Ismael Balafrej, Yann Beilliard, Dominique Drouin, Fabien Alibart, and Jean Rouat. Signals to spikes for neuromorphic regulated reservoir computing and emg hand gesture recognition. In International conference on neuromorphic systems 2021 , pages 1–8, 2021

    work page 2021

  10. [10]

    Signals to Spikes for Neuromorphic Regulated Reservoir Computing and EMG Hand Gesture Recognition

    Nikhil Garg, Ismael Balafrej, Yann Beilliard, Dominique Drouin, Fabien Alibart, and Jean Rouat. Signals to Spikes for Neuromorphic Regulated Reservoir Computing and EMG Hand Gesture Recognition. In International Conference on Neuromorphic Systems 2021 . Association for Computing Machinery, 2021

    work page 2021

  11. [11]

    Gokce, Abhishek K

    Elif I. Gokce, Abhishek K. Shrivastava, and Yu Ding. Fault tolerance analysis of surveil- lance sensor systems. IEEE Transactions on Reliability, 62(2):478–489, 2013

    work page 2013

  12. [12]

    Past, present, and future of sensor-based human activity recognition using wearables: A surveying tutorial on a still challenging task

    Harish Haresamudram, Chi Ian Tang, Sungho Suh, Paul Lukowicz, and Thomas Pl¨ otz. Past, present, and future of sensor-based human activity recognition using wearables: A surveying tutorial on a still challenging task. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 9(2), 2025

    work page 2025

  13. [13]

    Last-pain: Learning adaptive spike thresholds for low back pain biosignals classification

    Freek Hens, Mohammad Mahdi Dehshibi, Leila Bagheriye, Ana Tajadura-Jim´ enez, and Mahyar Shahsavari. Last-pain: Learning adaptive spike thresholds for low back pain biosignals classification. IEEE Transactions on Neural Systems and Rehabilitation Engi- neering, 33:1038–1047, 2025

    work page 2025

  14. [14]

    Tin Kam Ho and M. Basu. Complexity measures of supervised classification problems. IEEE Transactions on Pattern Analysis and Machine Intelligence , 24(3):289–300, 2002

    work page 2002

  15. [15]

    Evolving Comprehensive Prox- ies for Zero-Shot Neural Architecture Search, pages 1246–1254

    Junhao Huang, Bing Xue, Yanan Sun, and Mengjie Zhang. Evolving Comprehensive Prox- ies for Zero-Shot Neural Architecture Search, pages 1246–1254. Association for Computing Machinery, 2025

    work page 2025

  16. [16]

    Problems with health information technol- ogy and their effects on care delivery and patient outcomes: a systematic review

    Mi Ok Kim, Enrico Coiera, and Farah Magrabi. Problems with health information technol- ogy and their effects on care delivery and patient outcomes: a systematic review. Journal of the American Medical Informatics Association , 24(2):246–250, 2017

    work page 2017

  17. [17]

    Younghyun Kim, Woosuk Lee, Anand Raghunathan, Vijay Raghunathan, and Niraj K. Jha. Chapter 8 - reliability and security of implantable and wearable medical devices. In Swarup Bhunia, Steve J.A. Majerus, and Mohamad Sawan, editors, Implantable Biomed- ical Microsystems, pages 167–199. William Andrew Publishing, 2015

    work page 2015

  18. [18]

    Digital Twin in Healthcare: Patient System Mod- elling for Rehabilitation by Exoskeleton

    Martin Wolfgang Lauer-Schmaltz. Digital Twin in Healthcare: Patient System Mod- elling for Rehabilitation by Exoskeleton . Phd thesis, Technical University of Denmark, September 2024. Available at https://backend.orbit.dtu.dk/ws/files/394475134/ Lauer-Schmaltz_Digital_Twin_in_Healthcare.pdf

    work page arXiv 2024

  19. [19]

    AZ-NAS: Assembling Zero-Cost Proxies for Network Architecture Search

    Junghyup Lee. AZ-NAS: Assembling Zero-Cost Proxies for Network Architecture Search. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR), pages 5893–5903. IEEE, 2024

    work page 2024

  20. [20]

    Chapter 30 - Cognitive Walkthroughs

    Clayton Lewis and Cathleen Wharton. Chapter 30 - Cognitive Walkthroughs. In Mar- ting G. Helander, Thomas K. Landauer, and Prasad V. Prabhu, editors, Handbook of Human-Computer Interaction (Second Edition) , pages 717–732. North-Holland, second edition edition, 1997

    work page 1997

  21. [21]

    An Evaluation of Zero-Cost Prox- ies - from Neural Architecture Performance Prediction to Model Robustness.International Journal of Computer Vision , 133(5):2635–2652, 2025

    Jovita Lukasik, Michael Moeller, and Margret Keuper. An Evaluation of Zero-Cost Prox- ies - from Neural Architecture Performance Prediction to Model Robustness.International Journal of Computer Vision , 133(5):2635–2652, 2025

    work page 2025

  22. [22]

    Lyons and Wouter Vanderkulk

    Richard E. Lyons and Wouter Vanderkulk. The Use of Triple-Modular Redundancy to Improve Computer Reliability. IBM Journal of Research and Development, 6(2):200–209, 1962

    work page 1962

  23. [23]

    Emg-based gestures classification using a mixed-signal neuromorphic process- ing system

    Yongqiang Ma, Badong Chen, Pengju Ren, Nanning Zheng, Giacomo Indiveri, and Elisa Donati. Emg-based gestures classification using a mixed-signal neuromorphic process- ing system. IEEE Journal on Emerging and Selected Topics in Circuits and Systems , 10(4):578–587, 2020

    work page 2020

  24. [24]

    Neuromorphic implementation of a recurrent neural network for emg classifica- tion

    Yongqiang Ma, Elisa Donati, Badong Chen, Pengju Ren, Nanning Zheng, and Giacomo Indiveri. Neuromorphic implementation of a recurrent neural network for emg classifica- tion. In 2020 2nd IEEE International Conference on Artificial Intelligence Circuits and 13 Systems (AICAS), pages 69–73. IEEE, 2020

    work page 2020

  25. [25]

    Matthews

    B.W. Matthews. Comparison of the predicted and observed secondary structure of t4 phage lysozyme. Biochimica et Biophysica Acta (BBA) - Protein Structure , 405(2):442– 451, 1975

    work page 1975

  26. [26]

    Neural Architecture Search without Training

    Joe Mellor, Jack Turner, Amos Storkey, and Elliot J Crowley. Neural Architecture Search without Training. In Proceedings of the 38th International Conference on Machine Learn- ing, pages 7588–7598. PMLR, 2021

    work page 2021

  27. [27]

    Enhancing the explanatory power of usability heuristics

    Jakob Nielsen. Enhancing the explanatory power of usability heuristics. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems , pages 152–158. Association for Computing Machinery, 1994

    work page 1994

  28. [28]

    Evaluating emg feature and classifier selection for application to partial- hand prosthesis control

    Angkoon Phinyomark, Rami N Khushaba, Honghai Hu, Pornchai Phukpattaranont, and Chusak Limsakul. Evaluating emg feature and classifier selection for application to partial- hand prosthesis control. Frontiers in Neurorobotics, 6:17, 2012

    work page 2012

  29. [29]

    Counting the number of cells in immunocytochemical images using genetic algorithm

    Marjan Ramin, Alireza Sepas-Moghaddam, Payam Ahmadvand, and Mohammad Mahdi Dehshibi. Counting the number of cells in immunocytochemical images using genetic algorithm. In 12th International Conference on Hybrid Intelligent Systems (HIS) , pages 185–190. IEEE, 2012

    work page 2012

  30. [30]

    Robust Part-Based Hand Gesture Recognition Using Kinect Sensor

    Zhou Ren, Junsong Yuan, Jingjing Meng, and Zhengyou Zhang. Robust Part-Based Hand Gesture Recognition Using Kinect Sensor. IEEE Transactions on Multimedia, 15(5):1110– 1120, 2013

    work page 2013

  31. [31]

    Handbook of usability testing: How to plan, design, and conduct effective tests

    Jeffrey Rubin and Dana Chisnell. Handbook of usability testing: How to plan, design, and conduct effective tests. John Wiley & Sons, 2008

    work page 2008

  32. [32]

    Unintended Consequences of Wearable Sensor Use in Healthcare

    Michel Schukat, David McCaldin, Kejia Wang, Guenter Schreier, Nigel H Lovell, Michael Marschollek, and Stephen J Redmond. Unintended Consequences of Wearable Sensor Use in Healthcare. Yearbook of Medical Informatics, 25(01):73–86, 2016

    work page 2016

  33. [33]

    A novel hybrid algorithm for optimization in multimodal dynamic environ- ments

    Alireza Sepas-Moghaddam, Alireza Arabshahi, Danial Yazdani, and Mohammad Mahdi Dehshibi. A novel hybrid algorithm for optimization in multimodal dynamic environ- ments. In 12th International Conference on Hybrid Intelligent Systems (HIS) , pages 143–148. IEEE, 2012

    work page 2012

  34. [34]

    On the Prevalence of Sensor Faults in Real-World Deployments

    Abhishek Sharma, Leana Golubchik, and Ramesh Govindan. On the Prevalence of Sensor Faults in Real-World Deployments. In 2007 4th Annual IEEE Communications Society Conference on Sensor, Mesh and Ad Hoc Communications and Networks , pages 213–222, 2007

    work page 2007

  35. [35]

    Introduction to the Special Issue on Physiological Computing for Human-Computer Interaction

    Hugo Pl´ acido Da Silva, Stephen Fairclough, Andreas Holzinger, Robert Jacob, and Desney Tan. Introduction to the Special Issue on Physiological Computing for Human-Computer Interaction. ACM Trans. Comput.-Hum. Interact., 21(6), 2015

    work page 2015

  36. [36]

    The Design of Memristive Circuit for Affective Multi-Associative Learning

    Zilu Wang, Xiaoping Wang, Zezao Lu, Weiguo Wu, and Zhigang Zeng. The Design of Memristive Circuit for Affective Multi-Associative Learning. IEEE Transactions on Biomedical Circuits and Systems , 14(2):173–185, 2020

    work page 2020

  37. [37]

    A multimodal multilevel converged atten- tion network for hand gesture recognition based on semg and a-mode ultrasound

    Sheng Wei, Yue Zhang, and Honghai Liu. A multimodal multilevel converged atten- tion network for hand gesture recognition based on semg and a-mode ultrasound. IEEE Transactions on Cybernetics, 53(12):7723–7734, 2023

    work page 2023

  38. [38]

    Intelligent Wearable Systems: Opportu- nities and Challenges in Health and Sports

    Luyao Yang, Osama Amin, and Basem Shihada. Intelligent Wearable Systems: Opportu- nities and Challenges in Health and Sports. ACM Comput. Surv. , 56(7), 2024

    work page 2024

  39. [39]

    Neural Architecture Search with Reinforcement Learning

    Barret Zoph and Quoc Le. Neural Architecture Search with Reinforcement Learning. In International Conference on Learning Representations, 2017. 14

    work page 2017