Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Revisiting Noise Resilience Strategies in Gesture Recognition: Short-Term Enhancement in Surface Electromyographic Signal Analysis

Weiyu Guo, Ziyue Qiao, Ying Sun, Hui Xiong

TL;DR

This work tackles the challenge of noise resilience in surface EMG gesture recognition by prioritizing short-term information. It introduces STEM, a learnable module that captures local sEMG variations via sliding-window attention, and STET, which fuses long-term context with short-term signals through decoupled decoders and a fusion mechanism. A self-supervised pretraining regime (EIPC) with per-sensor masking and an asymmetric optimization objective further strengthens robustness and discrimination. Evaluations on GRABMyo and Ninapro DB2 show STET achieving superior accuracy and noise robustness, reducing the impact of noise by over 20% relative to strong baselines and demonstrating generalization across gesture recognition and regression tasks with practical deployment potential.

Abstract

Gesture recognition based on surface electromyography (sEMG) has been gaining importance in many 3D Interactive Scenes. However, sEMG is easily influenced by various forms of noise in real-world environments, leading to challenges in providing long-term stable interactions through sEMG. Existing methods often struggle to enhance model noise resilience through various predefined data augmentation techniques. In this work, we revisit the problem from a short term enhancement perspective to improve precision and robustness against various common noisy scenarios with learnable denoise using sEMG intrinsic pattern information and sliding-window attention. We propose a Short Term Enhancement Module(STEM) which can be easily integrated with various models. STEM offers several benefits: 1) Learnable denoise, enabling noise reduction without manual data augmentation; 2) Scalability, adaptable to various models; and 3) Cost-effectiveness, achieving short-term enhancement through minimal weight-sharing in an efficient attention mechanism. In particular, we incorporate STEM into a transformer, creating the Short Term Enhanced Transformer (STET). Compared with best-competing approaches, the impact of noise on STET is reduced by more than 20%. We also report promising results on both classification and regression datasets and demonstrate that STEM generalizes across different gesture recognition tasks.

Revisiting Noise Resilience Strategies in Gesture Recognition: Short-Term Enhancement in Surface Electromyographic Signal Analysis

TL;DR

This work tackles the challenge of noise resilience in surface EMG gesture recognition by prioritizing short-term information. It introduces STEM, a learnable module that captures local sEMG variations via sliding-window attention, and STET, which fuses long-term context with short-term signals through decoupled decoders and a fusion mechanism. A self-supervised pretraining regime (EIPC) with per-sensor masking and an asymmetric optimization objective further strengthens robustness and discrimination. Evaluations on GRABMyo and Ninapro DB2 show STET achieving superior accuracy and noise robustness, reducing the impact of noise by over 20% relative to strong baselines and demonstrating generalization across gesture recognition and regression tasks with practical deployment potential.

Abstract

Gesture recognition based on surface electromyography (sEMG) has been gaining importance in many 3D Interactive Scenes. However, sEMG is easily influenced by various forms of noise in real-world environments, leading to challenges in providing long-term stable interactions through sEMG. Existing methods often struggle to enhance model noise resilience through various predefined data augmentation techniques. In this work, we revisit the problem from a short term enhancement perspective to improve precision and robustness against various common noisy scenarios with learnable denoise using sEMG intrinsic pattern information and sliding-window attention. We propose a Short Term Enhancement Module(STEM) which can be easily integrated with various models. STEM offers several benefits: 1) Learnable denoise, enabling noise reduction without manual data augmentation; 2) Scalability, adaptable to various models; and 3) Cost-effectiveness, achieving short-term enhancement through minimal weight-sharing in an efficient attention mechanism. In particular, we incorporate STEM into a transformer, creating the Short Term Enhanced Transformer (STET). Compared with best-competing approaches, the impact of noise on STET is reduced by more than 20%. We also report promising results on both classification and regression datasets and demonstrate that STEM generalizes across different gesture recognition tasks.
Paper Structure (24 sections, 9 equations, 4 figures, 4 tables, 1 algorithm)

This paper contains 24 sections, 9 equations, 4 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. RELATED WORK
  3. Preliminaries
  4. Dataset
  5. Technology Detail
  6. Model Overview
  7. sEMG Intrinsic Pattern Capture Module
  8. sEMG Signal Encoding
  9. sEMG Signal Masking
  10. Long-term and Short-term Decoding
  11. Preserving Long-term sEMG Signal
  12. Preserving Short-term sEMG Signal
  13. Fusion
  14. Asymmetric Optimization
  15. Experiment
  16. ...and 9 more sections

Figures (4)

  • Figure 1: Schematic Diagram of EMG-based Human-Computer Interaction System.
  • Figure 2: Overview of STET. The sEMG signal is encoded using the sEMG Intrinsic Pattern Capture module, which is first pre-trained via sEMG signal Masking. A long-term and short-term enhanced module improves sEMG representations. An asymmetric optimization strategy addresses biases and imbalances in gesture recognition through an asymmetric classification loss.
  • Figure 3: Accuracy versus Noise Intensity Curve.
  • Figure 4: Visualization of (a) the long-term sEMG embeddings, (b) the short-term sEMG embeddings, and (c) the fused sEMG embeddings for gesture recognition. Note that we color each sample by its classes.

Theorems & Definitions (1)

  • Definition 1: sEMG Signal Sequence