A Framework and Call to Action for the Future Development of EMG-Based Input in HCI

Title of the Paper

A Framework and Call to Action for the Future Development of EMG-Based Input in HCI

Paper Information

• Subject Area: Human-Computer Interaction (HCI) and Electromyography (EMG) Input Technology
• Keywords: EMG technology, electromyography, dynamic gestures, static contraction, EMG control, design framework

Research Background and Problem Statement

• Identified Problems/Challenges:

  1. Electromyography (EMG) input has a long history in prosthetic control but remains underdeveloped for everyday interaction systems due to issues such as lack of robustness and intuitiveness.
  2. Current HCI research often directly adopts methods from prosthetics research without adequately addressing the unique requirements of HCI scenarios, such as the temporal structure in discrete input and dynamic gesture recognition.
  3. Issues such as poor interaction design, insufficient model development, lack of evaluation systems, and low reproducibility in research persist.

• Significance:
  EMG input is characterized by its "miniaturization," "low power consumption," and "non-invasiveness," making it well-suited to address gesture recognition and other non-mouse/keyboard interaction needs. However, its potential remains underexplored, particularly in general-purpose applications.

• Research Motivation and Related Work:
  While EMG signals are commonly used in prosthetics control, their broader adoption is hindered by the complexity of current methods and signal instability. This paper emphasizes that a design framework tailored to HCI and optimized input design are critical steps for the widespread adoption of EMG technology in human-computer interaction.

Proposed Solution

• Proposed Method/Solution:
  The authors summarize four major limitations of existing EMG technology and propose a design framework for EMG-based input specifically tailored to the HCI field. The framework includes the following eight key steps:

  1. Interaction Design: Define interaction types, including static contraction and dynamic gestures.
  2. Input Technology Selection: Consider sampling rate, number of electrodes, sensor type, etc.
  3. Control Scheme Design: Differentiate between continuous control and discrete control input schemes.
  4. Algorithm Selection: Use "static classification models" for static input and "temporal models" for dynamic gestures.
  5. Recognition Type Definition: Choose between "closed-set methods" (recognizing predefined gestures) and "open-set methods" (adapting to undefined gestures).
  6. Control Parameter Optimization: Adjust data segmentation strategies such as window size and stride.
  7. Model Training: Consider user dependency and context independence.
  8. System Evaluation: Emphasize online evaluation with real user participation.

• Novelty of the Solution:
  The framework provides a detailed and systematic approach to EMG interaction design from an HCI perspective, redefining the application process of EMG technology. This enables HCI researchers to design and evaluate EMG-based input systems more effectively.

• Key Technologies and Implementation Steps:

  • Interaction Type Classification: Use static contraction for continuous control signals and dynamic gestures for discrete event-based input design.
  • Temporal/Static Classification Model Matching Strategy: Select models such as Support Vector Machines (SVM) or Linear Discriminant Analysis (LDA) based on signal characteristics.
  • Parameter Optimization: Includes techniques for windowed segmentation, signal cleaning, and feature extraction.

Research Outcomes

• Specific Achievements:

  1. Proposed a design framework for EMG interaction systems tailored to HCI, addressing the need for robustness and complexity in general-purpose applications.
  2. Identified key issues and priority directions for future research, such as open-set recognition models, user training support, and activity detection.

• Advantages Over Existing Solutions:

  • Improved standardization for designing and reusing EMG input technology in the HCI community.
  • Specifically addressed the limitations of systems optimized for prosthetic control in meeting HCI needs.

• Experimental or Evaluation Results:
  The authors demonstrated the framework's practicality through various use cases (e.g., remote control of RC cars, music control during running), customizing experimental design processes for different HCI scenarios.

• Limitations and Future Directions:

  • The current framework does not encompass all complexities of EMG signals, such as signal variations caused by user fatigue or noise reduction in dynamic conditions.
  • Future directions include:
    1. Device Upgrades: Encourage the development of low-cost, durable, wearable EMG devices.
    2. Standardization of Discrete Gestures: Seek standardization and robustness in gesture input.
    3. Sensor Fusion-Based Methods: Explore hybrid input models combining EMG with inertial measurement units (IMUs).
    4. Open-Source Online Resources: Promote open-source datasets and code to enhance reproducibility and cross-disciplinary research.

Conclusion

This paper systematically reviews the functional potential and design requirements of EMG input technology in HCI. By proposing a clear framework and action plan, it provides significant guidance for future researchers, covering practical needs, best practices, and future development trends.