Related Experiment Video
Updated: Jul 15, 2025
09:28
Engineering Molecular Recognition with Bio-mimetic Polymers on Single Walled Carbon Nanotubes
Published on: January 10, 2017
8.2K
Convolution Neural Networks for Motion Detection with Electrospun Reversibly-Cross-linkable Polymers and Encapsulated
Su Bin Choi1, Hyun Sik Shin1, Jong-Woong Kim1,2
1Department of Smart Fab Technology, Sungkyunkwan University, Suwon 16419, Republic of Korea.
ACS Applied Materials & Interfaces
|October 2, 2023
Summary
A new polybutadiene-based urethane (PBU)/AgNW/PBU sensor (PAPS) offers superior mechanical stability for precise motion detection. Machine learning algorithms achieved over 98% accuracy in classifying motion signals from this advanced wearable sensor.
Area of Science:
- Materials Science: Development of novel composite materials for sensor applications.
- Engineering: Design and fabrication of advanced sensor technology.
- Computer Science: Application of machine learning for signal processing.
Background:
- Conventional sensors often lack mechanical robustness and are sensitive to bending.
- Existing fabrication methods for composite sensors can be complex and lead to imperfections.
- Need for reliable and durable sensors for accurate motion detection in wearable applications.
Purpose of the Study:
- To design, fabricate, and implement a novel polybutadiene-based urethane (PBU)/AgNW/PBU sensor (PAPS) with enhanced mechanical stability and motion detection capabilities.
- To investigate the performance of the PAPS compared to conventional AgNW/PBU sensors (APS).
- To explore the application of machine learning and deep learning algorithms for interpreting sensor signals.
Main Methods:
- Fabrication of the PAPS using electrospinning of PBU, integration of Ag nanowire (AgNW) electrodes, and sequential Diels-Alder (DA) and retro-DA reactions for encapsulation.
- Characterization of sensor performance, including mechanical stability, bending insensitivity, and signal generation from body motion.
- Application of K-means clustering for reproducibility assessment and deep learning models (1D CNN, LSTM) for signal classification.
Main Results:
- The PAPS demonstrated superior mechanical stability and bending insensitivity compared to APS.
- Distinctive signal curves were generated by the PAPS corresponding to specific body parts and motion degrees.
- K-means clustering confirmed higher reproducibility for PAPS signals.
- Deep learning models achieved over 98% classification accuracy, with a singular 1D CNN model showing excellent performance.
Conclusions:
- The PAPS represents a significant advancement in intelligent motion sensing due to its robust design and fabrication.
- The sensor's ability to generate reliable signals and its compatibility with machine learning enhance its utility for motion tracking.
- This technology offers a promising platform for developing next-generation wearable motion sensors with high precision and durability.
