Electrospinner: Flexible fiber-based dual-mode coupled tactile sensor: contactless gesture perception and high-precision material recognition!

Background

With the advancement of technology and the increase in the demand for human-computer interaction, the importance of front sensors in the fields of intelligent robots, prostheses, wearable devices, etc. has increased significantly. However, traditional front sensors have problems such as function settings, signal interruption during multi-mode collaborative work, and high consumption. The development of efficient, low-consumption and resource-integrated collectors has become a major challenge.

Recently, Professor Wan Yanfen's team at Yunnan University published a paper titled "Bimodal Coupling Haptic Perceptron for Accurate Non Contact Gesture Perception and Material" in Adv Fiber Mater. This work proposed a dual-mode connected multifunctional resource collector based on a variety of flexible fibers. Combined with machine learning algorithms, it achieved 5 types of non-contact auditory effective recognition with an accuracy rate of 98.5%; 10 different This study not only fundamentally improved the accuracy of speech recognition and material recognition, but also expanded the application prospects in smart devices.

The main point of this paper

Based on the rapid and slow adaptation of the human tactile perception system to external stimuli, this work proposes a capacitive and triboelectric dual-mode coupled tactile sensor to simulate the adaptability of human skin. The sensor consists of a capacitive sensor and a triboelectric sensor symmetrically distributed at both ends of polyethylene terephthalate (PET) to prevent signal crosstalk and reduce power consumption. In order to further broaden the potential application scenarios of the sensor, the sensor is used for non-contact gesture perception, including five gestures: single click, double click, tick, cross, and sweep, with a recognition accuracy of 98.5%. In addition, with the assistance of machine learning, the sensor also achieved accurate recognition of 10 different materials with a recognition accuracy of 100%

Among them, the capacitive sensor, as a pressure sensor, is composed of a VHB tape encapsulation layer, a hemispherical PDMS, polypyrrole, a thermoplastic polyurethane composite material (PPy@TPU), and a silicon paper dielectric layer. The sensor exhibits stability and linear response characteristics in the pressure range of 0 to 745.3 kPa, with a response time of 50 ms, and has good low-pressure detection capability and cyclic stability.

The triboelectric sensor is designed to simulate the non-contact perception ability of the human body. It uses thermoplastic polyurethane (TPU) nanofibers as the triboelectric layer, PPy@TPU as the electrode, and VHB tape as the insulating layer. The non-contact perception mechanism of the sensor is verified by polytetrafluoroethylene (PTFE) and nylon 6 (PA66) films with opposite polarities. In addition, for PTFE and PA66 of different polarities, the non-contact detection distances are 80 cm and 40 cm, respectively, laying the foundation for non-contact gesture recognition.

The triboelectric sensor is used for non-contact gesture recognition. By collecting the voltage signals of 5 gestures (single click, double click, outline, cross, sweep), using principal component analysis (PCA) and 6 different machine learning algorithm model training, the recognition accuracy of the random forest algorithm reached 96.3%. Therefore, a real-time gesture recognition system was developed based on the random forest algorithm, and non-contact gesture recognition was realized, showing great application prospects in fields such as virtual reality.

In addition, the dual-mode coupled sensor combining capacitive sensors and triboelectric sensors demonstrated the ability to identify materials with high sensitivity. Capacitive sensors and triboelectric sensors are responsible for capturing the hardness and electron affinity of different target materials, respectively, to distinguish and identify materials

To further develop the application scenarios of the sensor, the random forest algorithm was used for machine learning model training, achieving 100% accurate identification of 10 different materials, providing a solution for modern industrial automation and quality control, and potentially promoting intelligent manufacturing and efficient energy utilization.

Electrospinning Nanofibers Article Source:

https://link.springer.com/article/10.1007/s42765-024-00458-w