In 2024, by August, 50,133 Brazilians died from strokes (SBAVC, 2025). Additionally, from 2008 to 2015, over 361,000 upper and lower limb amputation surgeries were performed in Brazil (PEIXOTO, Alberto Monteiro et al., 2017). The absence or deficiency of limbs in these individuals leads to biomechanical adaptations to compensate for the loss (WILLIAMS et al., 2016). Thus, prostheses and orthoses play a fundamental role in rehabilitation.
This project focuses on the classification of movements using electromyographic (EMG) signals from the arm. The primary objective is to develop a model capable of identifying different movements based on myoelectric signals, contributing to the advancement of rehabilitation and assistive technologies.
A video summarizing the work can be found here
All the work development is available in the file main.ipynb, including:
- Data Processing: Preparation of EMG signals.
- Machine Learning Models: Use of multiple classifiers, including boosting and bagging techniques.
- Evaluation Metrics: Confusion matrix and classification reports.
- Visualization: Animated signal plots, real-time classification visualization, and segmentation of movement states.
The project outputs classification metrics, including precision, recall, F1-score, and an overall performance summary of different machine learning models. Additionally, the animated visualization provides an intuitive way to analyze movement classification in real-time.
If you have any questions or suggestions, feel free to open an issue or get in touch! 🚀