Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Freek Hens, Amirhossein Sadough, Aleksa Bokšan, Mahyar Shahsavari, Mohammad Mahdi Dehshibi

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.

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

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.

Paper Structure

This paper contains 23 sections, 6 equations, 5 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Evaluating Interactive Systems in HCI
  4. Fault Tolerance: From Heavyweight Engineering to Lightweight HCI
  5. A Priori Performance Expectation: Bridging Machine Learning and HCI
  6. A Framework for A Priori Robustness Auditing
  7. Data Representation in a Latent Feature Space
  8. Core Analytic: Quantifying Distributional Separability
  9. Maximum Fisher Discriminant Ratio (F1)
  10. Volume of Overlapping Region (F2)
  11. Maximum Individual Feature Efficiency (F3)
  12. Stage 1: A Priori Task Complexity Analysis
  13. Stage 2: Sensor Criticality and FTC Assessment
  14. Case Study: Auditing Robustness in an sEMG-based Gesture Interface
  15. Case Study Context: The Roshambo Dataset
  16. ...and 8 more sections

Figures (5)

  • Figure 1: Fall back mechanism for sEMG-based gesture recognition.
  • Figure 2: (a) Conceptual illustration of class separability. The figure visualises the varying degrees of distributional overlap for three gesture classes. The simplified 1D distributions at the bottom show that 'Rock' and 'Paper' are well-separated (low overlap, high separability), while 'Paper' and 'Scissors' have a large overlapping region (shaded), indicating they are inherently more difficult to distinguish from each other based on the sensor data. (b) Conceptual illustration of distributional shift. This figure illustrates how sensor ablation affects the distribution of the baseline, i.e., the intact sensors.
  • Figure 3: Conceptual illustration of the data collection setup using the Myo armband with eight sEMG sensors, based on the description in Garg et al. garg2020signals. The original dataset does not specify the anatomical mapping for each sensor index.
  • Figure 4: Stage 1 Results: (a) Our framework's model-free analysis predicts that the 'paper' vs 'scissors' task is over $10\times$ more difficult (lower FDR) than the other pairs. (b) This prediction is validated by an MLP classifier, which achieves a demonstrably lower MCC score on the same pair.
  • Figure 5: Stage 2 Results: Per-class sensor criticality. A larger bar indicates a more critical sensor for that specific gesture. The analysis identifies both indispensable sensors (e.g., Sensor 2 for Paper) and redundant ones (e.g., Sensors 6, 7).