Z-Ring: Single-point Bio-impedance Sensing for Gesture, Touch, Object and User Recognition

Title of the Paper

Z-Ring: Single-Point Bio-Impedance Sensing for Gesture, Touch, Object and User Recognition

Paper Information

• Research Area: Human-Computer Interaction, Wearable Devices, Novel Sensing Technologies
• Keywords: Sensing, Interaction, Bio-Impedance, Ring, Gesture Recognition, Object Recognition, User Identification, User Interface, Passive User Interface

Research Background and Problem

• Problems and Challenges: Current interaction technologies for wearable devices face limitations in sensing accuracy, functional diversity, and portability. Many methods require complex hardware, multi-point instrumentation, or environmental modifications, such as altering objects or interaction interfaces.
• Significance: Enhancing interaction capabilities for wearable devices can provide richer user experiences in fields such as gaming, augmented/virtual reality (AR/VR), and ubiquitous computing. A portable and multifunctional solution is needed.
• Motivation and Related Work:
  • Previous studies often focused on interactions achieved through environmental or hand-mounted devices (e.g., doorknobs, handles), but these methods require object modifications or complex instrumentation (e.g., cameras, sensor networks).
  • Z-Ring innovatively utilizes bio-impedance as a sensing method, relying solely on single-point instrumentation to detect gestures, object holding, and user identification through the electromagnetic properties of the hand.

Solution

• Main Approach: A single-point sensing device based on electric field excitation and bio-impedance detection—Z-Ring—is proposed. This wearable ring detects gestures, touch, object recognition, and user identification through changes in finger resistance.
• Innovations:
  • Single-point instrumentation without requiring modifications to the environment or interaction interface.
  • Rich interaction information obtained through broadband impedance detection, achieving greater functional diversity compared to previous devices.
  • Utilizing the human body as an antenna to detect changes in the electromagnetic properties of the hand.
• Implementation Steps and Techniques:
  1. Hardware Design: Z-Ring includes two electrodes—a signal electrode (for transmitting and receiving signals) and a bias electrode (for providing ground connection).
  2. Data Collection: A vector network analyzer (VNA) is used for frequency scanning, analyzing the S11 parameter (reflection coefficient) of the reflected signal to quantify hand states.
  3. Software Processing: Machine learning models (e.g., Support Vector Machine, Random Forest) are applied to analyze spectral data for interaction tasks.

Research Outcomes

• Specific Results:
  • Gesture Recognition: Capable of recognizing 5 types of single-hand and two-hand gestures. User-independent model accuracy reached 88% (single-hand) and 84% (two-hand), while user-dependent model accuracy reached 93% and 92%, respectively.
  • Passive UI Interaction: Developed battery-free buttons, sliders, and touchpads. Button recognition accuracy was 91.8%; 1D position prediction error for sliders was <4.4 cm, and 2D position prediction error for touchpads was <4.1 cm.
  • Object Recognition: Successfully recognized 6 common objects with an accuracy of 94.5%.
  • User Identification: Achieved user identification based on hand bio-impedance information with an accuracy of 99%, and user authentication accuracy of 98.3%.
• Advantages:
  • Does not rely on multi-point instrumentation or environmental modifications.
  • Bio-impedance enables lightweight, portable wearable devices capable of multitasking (e.g., gesture and object recognition).
  • Unified hardware design for multiple functionalities.
• Experiments and Evaluation: Conducted further experiments involving 21 participants, including tasks for gesture recognition, object recognition, UI interaction, and user identification, demonstrating high robustness and stability.
• Limitations and Future Directions:
  • Current hardware (VNA) is bulky and requires wired connections, limiting practical applications; future work should focus on optimizing for low-power, integrated sensing chips.
  • User-independent model accuracy needs further improvement, especially in object and button recognition tasks.
  • Explore how to enable mode switching for simultaneously running multiple application scenarios (e.g., gesture recognition and passive UI interaction) on the same device.

The above is a structured summary and analysis of the paper. Detailed experimental procedures and methods can be found in the appendix section at the end of the document.