IntelliLining: Activity Sensing through Textile Interlining Sensors Using TENGs

Research Background and Issues

• Identified Problems or Challenges: Traditional sensors are often integrated into clothing as aftermarket components. While this approach offers good functionality, it typically lacks aesthetic appeal and comfort and is difficult to seamlessly incorporate into garment manufacturing processes. In recent years, embedded sensing technology has gradually entered clothing design, but it primarily focuses on rigid components (e.g., buttons, zippers) and certain localized areas (e.g., drawstrings, seams). However, these solutions remain limited in sensing range and potential application scenarios.

• Importance: Utilizing clothing as an intelligent platform for activity sensing in daily life can provide new methods for health monitoring, skill assessment, and lifestyle tracking. In particular, clothing sensors that capture various activities of the hands, mouth, and body offer broad potential applications in these fields.

• Research Motivation and Related Work: Existing research has explored aftermarket sensors or smart clothing components, but most focus on recognizing single activities or specific types of garment parts. This paper introduces, for the first time, the concept of transforming the "lining" component of clothing into intelligent sensors, which provide multi-point sensing capabilities over a large area and can capture a wider range of activity data.

Solution

• Proposed Solution: The authors propose an intelligent lining system (IntelliLining) based on triboelectric nanogenerator (TENG) technology, which detects vibrations and activities. Unlike traditional sensing components, the intelligent lining offers a more distributed and integrated sensing layout, enabling activity detection from multiple areas of the garment.

• Innovations:

  1. Introduced the concept of embedding sensors into the lining component of clothing, which has been underexplored.
  2. Compared to existing solutions, the intelligent lining covers a larger sensing area through a two-dimensional surface layout.
  3. The intelligent lining can capture both airborne and surface vibration signals, exploring the effects of signal variations based on position, size, and vibration source.
  4. Capable of sensing diverse activity types (combinations of mouth, hand, and full-body activities), expanding the boundaries of wearable device sensing capabilities.

• Implementation Steps and Key Technologies:

  1. Sensor Fabrication: TENG sensors are constructed using a PTFE-coated film and two layers of conductive fabric, with friction between these layers generating electricity for vibration detection.
  2. Integration into Clothing: Sensors are embedded into lining positions such as cuffs, collars, and chest pockets using heat-pressing techniques, maintaining the garment's shape and comfort.
  3. Signal Collection and Analysis: Vibration signals are collected via data acquisition cards and classified using machine learning methods (including feature extraction-based approaches and deep learning networks) to identify activity types.
  4. Experimental Design and Data Collection: Investigated the impact of area size, deformation degree, vibration source location, and vibration transmission mode (airborne and surface propagation combined) on sensing performance.

Research Results

• Specific Findings:

  1. The intelligent lining sensors can reliably capture vibration signals from various daily activities, including airborne signals (e.g., speech) and surface vibrations (e.g., physical contact).
  2. The intelligent lining can recognize 12 common activities (e.g., writing, eating, walking, sneezing), achieving classification accuracy exceeding 93% when using deep learning models.
  3. The authors validated the stability of the sensors under different sizes, shapes, signal transmission media, and deformation scenarios.

• Advantages Over Existing Solutions: Compared to previous sensors based on single-point or small-area detection, the intelligent lining's primary advantage lies in its extensive sensing coverage, enabling parallel multi-point vibration data capture. Additionally, its two-dimensional layout is better suited for multimodal activity detection while retaining the garment's flexibility and wearing comfort.

• Experiments and Evaluation:

  • Experimental results demonstrate strong robustness of the sensors for activities of varying sizes and coverage locations.
  • Signal characteristics exhibit slight variations based on vibration source location, but these differences can be effectively addressed using machine learning.
  • Cross-validation in user experiments shows an activity recognition accuracy of approximately 85.5%, with potential for further improvement through expanded datasets.

• Limitations and Future Directions:

  1. Noise Handling: The current system is less effective in high-noise environments such as traffic noise; future work will incorporate adaptive noise reduction techniques.
  2. Material and Durability: Replace PTFE with more durable frictional fabric materials to enhance sensor longevity.
  3. User Experience and Integration: The current system is relatively bulky; future improvements will focus on lightweight embedded solutions to enhance wearability.
  4. Handling Undefined Activities: Employ anomaly detection techniques to better address untrained activities.

Conclusion

This paper introduces a novel intelligent lining technology based on TENG sensors, enabling precise sensing of multiple positions and activities, thereby expanding the capabilities of wearable devices. The preliminary results demonstrate high technical feasibility, opening new possibilities for innovative smart clothing design. Future improvements will focus on durability, noise handling, and optimization of embedded designs to bring the system closer to commercial applications.