Modeling Perceived Force of Electrical Muscle Stimulation to Improve User’s Recall

Paper Title

Modeling Perceived Force of Electrical Muscle Stimulation to Improve User’s Recall

Publication Info

• Topic area: Improving force recall in electrical muscle stimulation (EMS) interfaces for physical skill acquisition.
• Keywords: Electrical muscle stimulation, force recall, haptics, skill acquisition, user modeling, perceptual mismatch, regression models, force-feedback, human-computer interaction, motor skills.

Background and Problem

• Problem / challenge: Existing EMS interfaces are effective for demonstrating movements but fail to accurately demonstrate forces, leading to significant user recall errors. Users tend to overshoot the perceived force, especially at lower force levels, with a median overshoot of 19%.
• Significance: Accurate force demonstration is critical for skill acquisition in tasks requiring precise force application, such as playing musical instruments or operating tools. Without addressing this issue, EMS interfaces cannot fully realize their potential in human-computer interaction (HCI).
• Motivation and related work: Prior research focused on improving EMS trajectory and pose accuracy but neglected force perception and recall. Although EMS has been used in VR and AR for immersive interactions, these applications do not prioritize accurate force demonstration. This paper addresses the gap by modeling user perception of EMS-demonstrated forces to reduce recall errors.

Solution

• Proposed approach: Develop a perceptual model of user recall for EMS-demonstrated forces and use it to adjust target forces, improving recall accuracy.
• Novelty:
  1. Identification and quantification of the force overshoot problem in EMS interfaces.
  2. Development of personalized regression models to predict and adjust for perceptual mismatches in force recall.
  3. Validation of the approach through experiments showing a 35% reduction in recall error.
  4. Exploration of potential applications for force-focused EMS systems.
• Procedure and key techniques:
  1. Conduct a user study to measure force recall errors for EMS-demonstrated forces.
  2. Develop and evaluate three regression models (linear, quadratic, power law) for predicting user recall.
  3. Validate the best-performing personalized models by testing them on unseen target forces.
  4. Compare model-adjusted EMS demonstrations to baseline EMS demonstrations in terms of recall accuracy.

Results

• Concrete findings:
  • Median recall error was reduced by 35% when using model-adjusted EMS demonstrations compared to baseline EMS demonstrations.
  • Overshoot errors were most pronounced at lower force levels (e.g., 80.2% median error at 500g) and decreased at higher force levels.
  • Linear regression models were the most effective for 6 out of 12 participants, with constrained-quadratic and power models performing better for others.
• Advantage over baselines: Model-adjusted EMS demonstrations significantly outperformed baseline EMS demonstrations, particularly at lower and mid-force levels.
• Experiments / evaluation:
  • Phase 1: Measured recall errors for three target force levels (minimum, midpoint, maximum) across 30 trials per participant.
  • Phase 2: Developed personalized regression models using cross-validation and compared them to a global model.
  • Phase 3: Validated the models on 15 unseen target forces, showing significant improvements in recall accuracy.
• Limitations and future work:
  • Limited number of calibration trials for model generation.
  • Potential confounds from multiphase study design, though mitigated by randomized trial conditions.
  • Focus on simple regression models; future work could explore machine learning approaches.
  • Results may not generalize across larger populations due to individual variability in EMS sensitivity.

Summary

This study addresses the challenge of force recall errors in EMS interfaces by developing personalized perceptual models to adjust target forces. The approach reduced median recall errors by 35% and was particularly effective at lower and mid-force levels. The findings highlight the importance of modeling user perception for precise force demonstrations and open new possibilities for EMS applications in skill acquisition. Future research could explore more complex modeling techniques and extend the approach to dynamic and multi-modal EMS systems.