Assisting with Fingertip Force Control by Active Bio-Acoustic Sensing and Electrical Muscle Stimulation

Title of the Paper

Assisting with Fingertip Force Control by Active Bio-Acoustic Sensing and Electrical Muscle Stimulation

Paper Information

• Research Area: Human-Computer Interaction, Fingertip Force Control Assistance Based on Electrical Muscle Stimulation (EMS) and Active Bio-Acoustic Sensing
• Keywords: Electrical Muscle Stimulation, Active Bio-Acoustic Sensing, Closed-Loop Control, Fingertip Force, Pinch Force

Research Background and Problem

• Problem:
  • Fingertip force control is crucial for learning motor skills, such as in golf, surgical operations, and other precision tasks.
  • Existing solutions, such as exoskeleton gloves, cover the fingers and interfere with users' finger movement control and tactile feedback.
  • Current solutions using surface electromyography (EMG) sensors do not require finger-mounted devices but are susceptible to noise from EMS electrical signals.
• Significance:
  • Focuses on fingertip force assistance devices that do not hinder hand flexibility or tactile feedback, aiding in the enhancement of motor skills.
• Research Motivation and Related Work:
  • Existing studies primarily focus on guiding users' posture and movements through EMS, with limited exploration of closed-loop control for fingertip force.
  • Active Bio-Acoustic Sensing (ABAS) is a relatively new sensing technology capable of distinguishing different hand movements and force levels.

Solution

• Method and Principle:
  • Proposed a wearable system that estimates fingertip force through active bio-acoustic sensing and incorporates electrical muscle stimulation (EMS) for unconscious force control.
  • The system uses piezoelectric elements (located on the back of the hand) for force estimation and EMS (applied to the forearm) for feedback in force control.
• Innovations:
  • The system does not require devices to be worn on the fingers, preserving finger flexibility and tactile integrity.
  • ABAS is insensitive to EMS electrical signals, enabling collaborative operation within the same system.
• Key Technologies and Implementation:
  1. Active Bio-Acoustic Sensing:
     • Utilizes two piezoelectric elements (one emitting sound waves and the other receiving vibration responses) to estimate pinch force.
     • Processes feature vectors using machine learning regression models (Support Vector Regression, SVR, or Gaussian Process Regression, GPR).
  2. Electrical Muscle Stimulation:
     • Employs Pulse Frequency Modulation (PFM) for closed-loop control of EMS intensity, stimulating target muscles (flexor pollicis longus, extensor pollicis longus).
     • Adjusts EMS using proportional control or switch control methods.
  3. Hardware Prototypes:
     • Designed three hardware prototypes: fixed with medical tape, embedded in a wristband frame, and integrated into a fingerless glove.
     • All three prototypes can estimate forces at different target levels.

Research Outcomes

• Specific Results:
  • The system effectively assists users in pinch force control at different target levels (e.g., light force 3N, medium force 6N) using EMS, significantly reducing errors.
  • The system does not require devices to cover the fingers, maintaining tactile feedback while achieving interactive and precise force control.
• Experimental and Evaluation Results:
  • For force estimation accuracy, the average error of the three prototypes ranged from 0.91 to 1.38N, with correlation coefficients of approximately 0.86 to 0.93.
  • In user studies, most participants showed significantly reduced force errors at low to medium target force levels.
  • Performance was limited in dynamic environments (e.g., golf putting tests) and at high target force levels (e.g., 9N), likely due to limitations in force estimation accuracy.
• Advantages Compared to Existing Solutions:
  • Compared to exoskeleton gloves or EMG-based methods, the system addresses issues of device interference with hand movements or tactile feedback.
  • The combination of EMS and ABAS enables closed-loop and real-time force control without the need to redesign force feedback hardware.
• Limitations and Future Directions:
  • Current force estimation is sensitive to motion artifacts, reducing accuracy in dynamic conditions.
  • The system requires extensive calibration for different environments and users, necessitating large datasets to improve model robustness.
  • Future plans include enhancing the estimation model with deep learning implementations, optimizing hardware portability, and extending applications to more complex fingertip force control scenarios (e.g., piano playing).