0100499781

open access

Version of Record

Flexible Optical Tactile Sensor With CNN-Based Calibration for Robust Pressure Estimation

English (英語)

IEEE Sensors Journal

26(3)

3896-3906

Institute of Electrical and Electronics Engineers (IEEE)

2026-02-01

2026-02-04

Accurately estimating pressure without interfering with natural tactile perception remains a critical challenge for wearable tactile sensors, whose outputs are often affected by attachment conditions and individual variability. Here, we present a flexible optical tactile sensor that estimates pressure from optical signals reflecting both skin deformation and hemodynamic changes. To enhance robustness and generalization across different placements and users, a convolutional neural network (CNN)-based calibration framework was developed. The sensor is less than 150 μm thick and features a perforated architecture that reduces perspiration effects while maintaining comfort and stable contact during long-term use. Experiments conducted on multiple fingers across a 0–100 kPa (0–4 N) range demonstrated stable and accurate pressure estimation, with consistent performance maintained after reattachment and across subjects. These results confirm that the CNN calibration effectively compensates for placement- and subject-dependent variability. Moreover, characteristic modulations in the optical response associated with blood-flow dynamics were observed under loading, indicating the capability to simultaneously capture mechanical and physiological information. Overall, the proposed optical tactile sensor with CNN-based calibration provides a promising platform for quantitative tactile sensing in wearable devices, human–robot interaction, and biomedical applications.

tactile

学術雑誌論文

This work is licensed under a Creative Commons Attribution 4.0 License