Abstract
Human-machine interaction has emerged as a revolutionary and transformative technology, bridging the gap between human and machine. Gesture recognition, capitalizing on the inherent dexterity of human hands, plays a crucial role in human-machine interaction. However, existing systems often struggle to meet user expectations in terms of comfort, wearability, and seamless daily integration. Here, we propose a handwriting recognition technology utilizing an intelligent hybrid-fabric wristband system. This system integrates spun-film sensors into textiles to form the smart fabric, enabling intelligent functionalities. A thermal encapsulation process is proposed to bond multiple spun-films without additional materials, ensuring the lightweight, breathability, and stretchability of the spun-film sensors. Furthermore, recognition algorithms facilitate precise accurate handwriting recognition of letters, with an accuracy of 96.63%. This system represents a significant step forward in the development of ergonomic and user-friendly wearable devices for enhanced human-machine interaction, particularly in the virtual world.
Introduction
Human-Machine Interaction (HMI) is rapidly evolving toward wearable, "invisible" interfaces. However, current solutions often face a difficult trade-off between performance and user comfort. Non-contact systems, such as cameras, are prone to environmental interference, while contact-based devices like data gloves are frequently bulky, restrictive, and lack breathability.
Wrist-based wearables offer a strategic middle ground, but capturing the subtle muscle and tendon movements required for fine motor tasks—like writing a letter in the air—is technically challenging. Most existing electronic textiles rely on adhesives to laminate sensors, which adds weight and reduces breathability. This study introduces an Intelligent Hybrid-Fabric Wristband (IHFW) system that bridges the gap between textile engineering and deep learning to provide a comfortable, high-precision interface.
Methods
The core innovation of the IHFW system lies in its manufacturing and signal processing approach:
- Thermal Encapsulation: To avoid the use of external adhesives that degrade comfort, the researchers utilized the inherent melting properties of electrospun thermoplastic polyurethane (TPU). By hot-pressing layers at an optimal temperature of 115°C, they fused the dielectric and electrode layers (TPU mixed with silver nanowires) into a single, integrated fabric. This "all-spun" structure preserves the material's porosity, ensuring the device remains lightweight, breathable, and stretchable.
- Capacitive Sensing Array: The wristband features a six-channel capacitive pressure sensor array. Unlike standard circuits that measure voltage, this system employs an NE555-based capacitance-to-frequency conversion. By translating pressure changes into frequency shifts, the system achieves significantly higher resistance to environmental noise and electromagnetic interference.
- 1D-Convolutional Neural Network (1D-CNN): To process the time-series data from the six channels, a 1D-CNN architecture was implemented. This model is specifically tuned to recognize patterns in motion data—such as the unique "tendon travel" associated with individual finger movements—without requiring the massive datasets typical of more complex architectures.
Key Results and Discussion
The IHFW system was evaluated through custom datasets involving air-writing of the 26 English letters and object-grasping tasks. Key performance metrics include:
- High Recognition Accuracy: The system achieved a 96.63% accuracy for recognizing handwritten letters. It also reached 100% accuracy in identifying eight different grasped objects after only 15 training samples per object.
- Rapid Response: The sensors boast a response time of 12 ms and a recovery time of 18 ms, enabling seamless real-time interaction.
- Durability and Comfort: The thermal encapsulation allowed the device to survive 1,500 pressure cycles with minimal signal degradation. Strategic channel mapping—where specific sensors capture large wrist joint movements while others capture subtle finger movements—ensures high-fidelity data collection across various hand gestures.
Conclusion
The intelligent hybrid-fabric wristband demonstrates a successful integration of breathable electronic textiles with robust machine learning algorithms. While the current prototype requires individual user calibration and further testing for long-term wear (beyond 1,500 cycles), it represents a significant advancement in ergonomic HMI. Future work will focus on expanding the dataset to include more diverse handwriting styles and improving the long-term mechanical robustness of the silver nanowire-TPU interface.
Summary updated