.png){kind=icon}
.png){kind=icon}
Return
Adaptive Fuzzy Residual Learning Framework for On-Skin Triboelectric Sensor Gesture Recognition
.png){kind=icon}
PRE
AI
Z
H
Y
X
X
J
J
DOI：10.1109/tfuzz.2026.3695771.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This article proposes an adaptive fuzzy residual network (AFRN) for triboelectric signal-based gesture recognition that couples deep temporal feature learning with interpretable Takagi–Sugeno–Kang (TSK) fuzzy reasoning. Built on a temporally focused convolutional neural network backbone, AFRN introduces residual TSK inference blocks (ResTSK-Blocks) that embed learnable membership functions and rule consequents into the intermediate feature pipeline and optimize them end-to-end via backpropagation. The ResTSK-Block performs fuzzy inference at each temporal index and injects compact fuzzy semantics into the convolutional stream through a shape-aligned projection with residual fusion. A cross-layer feature fusion strategy further aggregates multidepth representations to improve discrimination. The learned rule activations and membership parameters are directly inspectable, enabling rule-level interpretability alongside deep feature learning. Experiments on a self-collected triboelectric signal dataset with 26 gesture classes from six subjects show that AFRN achieves an average accuracy of 99.82% under the subject-dependent protocol and outperforms representative baselines. Ablation studies further validate the contributions of ResTSK-Blocks and cross-layer feature fusion, confirming the effectiveness of integrating fuzzy reasoning with hierarchical deep feature learning.
Keywords:
Feature fusion
gesture recognition
human–machine interaction
Takagi–Sugeno–Kang (TSK) fuzzy inference
Journal
IF:
11.9
Papers:
4.9K
Citations:
2.9W
Organization
Cited Papers