C-DTW for Human Action Recognition Based on Nanogenerator.

Area of Science:

• Wearable sensor technology
• Biomedical engineering
• Human-computer interaction

Background:

• Human action recognition (HAR) systems often rely on wearable sensors to collect motion data.
• Existing HAR research prioritizes accuracy, often overlooking energy efficiency in portable systems.
• Triboelectric nanogenerators (TENGs) offer high sensitivity and energy harvesting potential for wearable sensors.

Purpose of the Study:

• To develop an energy-efficient wearable sensor for HAR using TENGs.
• To propose a novel algorithm, Compensatory Dynamic Time Warping (C-DTW), to enhance HAR classification accuracy.
• To evaluate the performance of C-DTW against existing Dynamic Time Warping (DTW) variants.

Main Methods:

• Fabrication of a TENG-based wearable sensor using polydimethylsiloxane and carbon nanotube film.
• Acquisition of sensor data using a sliding window approach for real-time identification.
• Development and simulation of the C-DTW algorithm to address quasi-periodic data challenges.

Main Results:

• The fabricated TENG sensor achieved an output performance of 9.98 mW/cm2.
• C-DTW demonstrated higher classification accuracy compared to DTW, WDTW, DDTW, and softDTW.
• C-DTW exhibited comparable complexity to DTW and was faster than shapeDTW for similar accuracy.

Conclusions:

• The TENG-based sensor is a viable component for energy-efficient HAR systems.
• C-DTW offers a significant improvement in classification accuracy for HAR, outperforming existing DTW methods.
• The compensatory mechanism of C-DTW can enhance the performance of various DTW-based HAR approaches.