PTCVS-net: Patch-based transformer-CNN with variable sparse attention network for sEMG-based gesture recognition

doi:10.1016/j.asoc.2025.114316

PTCVS-net: Patch-based transformer-CNN with variable sparse attention network for sEMG-based gesture recognition

2025-11-22

0

PRE

AI

S

Sike Ni

M

Mohammed A. A. Al‐qaness *

DOI：10.1016/j.asoc.2025.114316

Original

Original request for help

Save

Abstract

En 中文

• Propose a novel model for sEMG-based gesture recognition using an advanced deep learning structure. • Utilize Patch for fine-grained analysis of signals and capturing spatial correlation patterns between channels. • Propose a Transformer-CNN architecture to capture complementary time-frequency domain information and perform decision-level integration. • Apply diagonal sparse attention to dynamically select effective features and accurately focus on salient features. • Employ hierarchical extension heads to achieve high-resolution feature expression and enhance the ability to distinguish similar features.

PTCVS-net: Patch-based transformer-CNN with variable sparse attention network for sEMG-based gesture recognition

Abstract

Journal

Organization