Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Tackling Electrode Shift In Gesture Recognition with HD-EMG Electrode Subsets

Joao Pereira, Dimitrios Chalatsis, Balint Hodossy, Dario Farina

TL;DR

The paper addresses robustness of sEMG gesture recognition to electrode shift and session variability. It proposes training on HD-sEMG channel subsets with shift-based data augmentation to learn shift-invariant mappings while reducing input dimensionality. Using the Capgmyo dataset, the authors demonstrate that this AVS training improves intersession performance across feature sets and classifiers without relying on transfer learning. This approach offers a practical route to robust, cost-effective EMG-based interfaces and enables domain adaptation across different HD-sEMG acquisition setups.

Abstract

sEMG pattern recognition algorithms have been explored extensively in decoding movement intent, yet are known to be vulnerable to changing recording conditions, exhibiting significant drops in performance across subjects, and even across sessions. Multi-channel surface EMG, also referred to as high-density sEMG (HD-sEMG) systems, have been used to improve performance with the information collected through the use of additional electrodes. However, a lack of robustness is ever present due to limited datasets and the difficulties in addressing sources of variability, such as electrode placement. In this study, we propose training on a collection of input channel subsets and augmenting our training distribution with data from different electrode locations, simultaneously targeting electrode shift and reducing input dimensionality. Our method increases robustness against electrode shift and results in significantly higher intersession performance across subjects and classification algorithms.

Tackling Electrode Shift In Gesture Recognition with HD-EMG Electrode Subsets

TL;DR

The paper addresses robustness of sEMG gesture recognition to electrode shift and session variability. It proposes training on HD-sEMG channel subsets with shift-based data augmentation to learn shift-invariant mappings while reducing input dimensionality. Using the Capgmyo dataset, the authors demonstrate that this AVS training improves intersession performance across feature sets and classifiers without relying on transfer learning. This approach offers a practical route to robust, cost-effective EMG-based interfaces and enables domain adaptation across different HD-sEMG acquisition setups.

Abstract

sEMG pattern recognition algorithms have been explored extensively in decoding movement intent, yet are known to be vulnerable to changing recording conditions, exhibiting significant drops in performance across subjects, and even across sessions. Multi-channel surface EMG, also referred to as high-density sEMG (HD-sEMG) systems, have been used to improve performance with the information collected through the use of additional electrodes. However, a lack of robustness is ever present due to limited datasets and the difficulties in addressing sources of variability, such as electrode placement. In this study, we propose training on a collection of input channel subsets and augmenting our training distribution with data from different electrode locations, simultaneously targeting electrode shift and reducing input dimensionality. Our method increases robustness against electrode shift and results in significantly higher intersession performance across subjects and classification algorithms.
Paper Structure (9 sections, 2 figures, 2 tables)

This paper contains 9 sections, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. METHODS
  3. Dataset
  4. Electrode Subset Sampling
  5. Preprocessing & Machine Learning
  6. EXPERIMENTAL ANALYSIS
  7. Electrode Shift Generalization
  8. Intersession Performance
  9. CONCLUSION

Figures (2)

  • Figure 1: Schematic of channel subset sampling procedure applied to the Capgmyo dataset set-up. Three training conditions are considered: (i) training only on the central channel subset, where one channel per acquisition channel is considered; (ii) training/testing on all (128) channels; (iii) training on all valid channel subsets, where only proximal-distal translations are considered.
  • Figure 2: Architecture of our temporal VGGNet. The network consists of a number of TVGG blocks consisting of two temporal convolutions, a maxpool and spatial dropout. Each TVGG block doubles the number of channels, while keeping the temporal dimension constant.