uKnit: A Position-aware Reconfigurable Machine-knitted Wearable for Gestural Interaction and Passive Sensing using Electrical Impedance Tomography

Document Title

uKnit: A Position-Aware Reconfigurable Machine-Knitted Wearable for Gestural Interaction and Passive Sensing using Electrical Impedance Tomography

Document Information

• Subject Area: Wearable Devices, Human-Computer Interaction, Smart Textile Technology
• Keywords: Reconfigurable Wearable Devices, Smart Textiles, Knitting Technology, Electrical Impedance Tomography (EIT), Gestural Interaction

Research Background and Problem

• Identified Problems or Challenges:

  • Existing soft wearable devices typically have fixed forms and functions, limiting their adaptability to user needs.
  • Most smart textile products are single-function and cannot support multi-scenario applications.
  • Wearable devices lack the ability to differentiate and adapt to different body parts or multifunctional scenarios.

• Significance:

  • Enhancing the comfort and multifunctionality of wearable devices can meet the demands of broader applications.
  • Soft and lightweight textile technology has vast potential for everyday applications.

• Research Motivation and Related Work:

  • The authors were inspired by reconfigurable accessories (e.g., multifunctional scarves) to explore the potential of soft textile devices as multifunctional input and sensing channels.
  • In current research, Electrical Impedance Tomography (EIT) has been applied to gesture recognition and tactile sensing, but its exploration in the field of flexible smart fabrics is limited.

Solution

• Proposed Solution:

  • Designed and manufactured uKnit, a machine-knitted smart textile wearable device that supports Electrical Impedance Tomography.
  • Enabled multi-location usage of the device (e.g., head, arm, waist) with gesture recognition and passive sensing capabilities.

• Innovations:

  • Proposed a textile sensor capable of detecting body-wearing positions.
  • Combined EIT technology with flexible knitted sensors to achieve recognition and sensing of multi-posture gestures.
  • Incorporated machine learning models for wear position detection and gesture recognition.

• Implementation Steps:

  1. Designed sensing fabrics using industrial knitting machines and integrated electrical impedance sensing technology to create sensing modules.
  2. Developed model training and signal processing algorithms for position detection, gesture recognition, and passive sensing.
  3. Validated the device's effectiveness across various scenarios and tasks through user studies.

Research Outcomes

• Specific Results:

  • Position Detection: For single-user/multi-user models, uKnit achieved 88.0% and 78.2% accuracy, respectively, in 5-class wear position detection.
  • Gesture Recognition: For 7 gesture types, the single-user/multi-user models achieved accuracies of 80.4% and 75.4%, respectively.
  • Passive Sensing: Achieved a breathing rate recognition error rate of 1.25 breaths/min and a sitting posture detection accuracy of 86.2%.

• Advantages:

  • Provided a multifunctional soft wearable device with unified functionality, reducing dependency on multiple devices.
  • Combined configurability and comfort, making the device suitable for diverse scenarios and needs.

• Experimental or Evaluation Results:

  • User studies confirmed the device's adaptability to multiple scenarios and high accuracy performance.
  • Preliminary washability tests showed good mechanical integrity of the sensors after washing, though electrical performance was compromised.

• Limitations and Future Directions:

  • The durability of fabric materials and electrodes needs further optimization, especially for washing.
  • The design of mechanical and electronic connection components requires improved durability.
  • Future work could expand the device to other diverse forms (e.g., smart furniture accessories).
  • Explore reducing the number of electrodes to simplify the overall device design and enhance its versatility.