EMG-Based Classification of Standing Postures

This repository contains code for classifying four standing postures using EMG signals recorded from the soleus (SOL) and flexor digitorum brevis (FDB).

The primary objective is to compare how effectively each muscle’s activity can discriminate between standing postures.

---

Overview

A DeepConvNet-style convolutional neural network is used to classify posture from segmented EMG windows. The analysis includes:

• Model training and evaluation
• Posture-wise accuracy analysis
• Performance comparison between SOL and FDB

---

Scientific Motivation

Standing balance involves coordinated activation of intrinsic (FDB) and extrinsic (SOL) muscles. This project evaluates whether one muscle group provides more posture-specific EMG information.

---

Data

The original EMG data are not included due to IRB and study protocol restrictions.

Expected data format:

• Input: n × 64 × 512 EMG windows
• Labels: integers (1–4)

---

Repository Structure

emg-posture-classification/
├── notebooks/
│   └── EMG_Posture_Classification_GitHub.ipynb
├── src/
│   └── emg_posture_classification.py
├── requirements.txt
└── README.md

Notes

• Preprocessing and segmentation are assumed to be completed prior to loading
• Consider subject-wise splits to avoid data leakage
• Multiple runs are recommended for stable performance estimates

---

Citation

Code

If you use this repository, please cite:

Kamankesh, A. (2026). EMG-Based Classification of Standing Postures (Version 1.1) [Computer software]. Zenodo. https://doi.org/10.5281/zenodo.19825555

Paper

This repository accompanies the following publication:

Kamankesh, A., Rahimi, N., Amiridis, I. G., Sahinis, C., Hatzitaki, V., & Enoka, R. M. (2025).
Distinguishing the activity of flexor digitorum brevis and soleus across standing postures with deep learning models.
Gait & Posture, 117, 58–64. https://doi.org/10.1016/j.gaitpost.2024.12.014