Machine Learning-Enhanced Biomass Pressure Sensor with Embedded Wrinkle Structures Created by Surface Buckling.

Area of Science:

• Materials Science
• Biomedical Engineering
• Wearable Technology

Background:

• Flexible piezoresistive sensors are crucial for wearable devices but face challenges in natural material preparation and signal identification.
• Existing sensors often struggle with high precision and distinguishing similar, short vibration signals.

Purpose of the Study:

• To develop a high-precision, biomass-based flexible piezoresistive sensor using natural materials.
• To enhance the identification of subtle body movements and similar short vibration signals using machine learning.

Main Methods:

• Fabrication of flexible piezoresistive sensors using konjac glucomannan and k-carrageenan composite hydrogel with wrinkled surfaces.
• Coating wrinkled hydrogel surfaces with MXene sheets for enhanced sensitivity.
• Application of an XGBoost machine learning model for signal identification of throat vibrations.

Main Results:

• The sensor demonstrated high sensitivity (5.1 kPa⁻¹ at 50 Pa), a fast response time (104 ms), and excellent stability (>100 cycles).
• The XGBoost model successfully distinguished short voice vibrations corresponding to individual English letters.
• Experiments and simulations elucidated the mechanism behind wrinkle structure formation and sensing performance.

Conclusions:

• The developed biomass sensor offers a convenient and effective method for preparing high-precision flexible piezoresistive sensors.
• The integration of this sensor with machine learning significantly improves the detection and identification of subtle body movements and voice signals.
• This approach holds promise for advancing the design and application of next-generation wearable devices.