EMG Signal Processing - Movement Classification & Joint-Angle Regression NX-421 Project 2

This repository contains the code and report of our work on the NX-421 mini-project 2.

---

We perform a full pipeline—from raw electromyography (EMG) preprocessing to machine learning models for classifying hand/wrist movements and predicting joint angles using the NinaPro database. The work is divided into two main tasks:

1. Movement Classification

   • Preprocess EMG signals (filtering, envelope extraction, segmentation)
   • Extract time- and frequency-domain features
   • Train and optimize support-vector classifiers (SVC) via grid search
   • Evaluate performance using accuracy, confusion matrices and F₁ scores

2. Joint-Angle Regression

   • Process synchronized EMG and kinematic data
   • Select informative features with mutual information and PCA
   • Use multi-output support-vector regression (SVR)
   • Tune hyperparameters and assess with RMSE and MAE metrics

---

The code and environment.yml are provided for reproducibility. See the report for details about results and analysis.