Project Title

Team

Supervisors

Table of content

  1. Abstract
  2. Related works
  3. Methodology
  4. Experiment Setup and Implementation
  5. Results and Analysis
  6. Conclusion
  7. Publications
  8. Links

Abstract

Brain Computer Interfaces(BCIs) have made a huge impact in neurotechnology by enabling direct communication between the human brain and electronic external systems. Motor Imagery (MI) based BCIs, which utilize electroencephalography(EEG) signals to interpret motor imagery tasks have demonstrated significant potential in assistive technologies and neurorehabilitation. Traditional MI classification approaches typically distinguish between broad movements. However, novel research suggests that targeting single-joint MI EEG classification (Eg:Wrist, Elbow, Knee) can enhance precision, providing more intuitive and natural control for BCI applications. This study focuses on a single-joint, specifically the wrist, based MI task classification. The introduction of a novel model capable of distinguishing distinct wrist movements and investigating model performance across distinct joint movements, is the main aim of this study. Thus, contributing to improved precision in BCI applications.

Traditional MI classification research so far focuses on distinguishing between broad movement categories, such as left-hand and right-hand imagery.

Methodology

The primary objective is to classify MI tasks for single joint movements, focusing on the wrist, with high accuracy to enhance the usability of Brain-Computer Interface(BCI) applications in neurorehabilitation and assistive technology.

EEG Data

EEG signals of 25 healthy subjects performing wrist movements, pronation and supination, are obtained through the publicly available dataset on GigaDB(https://gigadb.org/dataset/100788).

Data Preprocessing

Data Preprocessing is required to be carried out in order to develop models based on traditional machine learning. Techniques proposed to be utilized would include,

Preprocessing Techniques

Feature Extraction

Feature Extraction

Classification Approaches

Traditional Machine Learning Models include,

Deep Learning Models include,

Classification

Model Evaluation

The model’s performance will be assessed using standard classification metrics, including accuracy, precision, recall, F1-score, and Matthews Correlation Coefficient (MCC). Additionally, Area Under the ROC Curve (AUC-ROC) will provide insights into the model’s discriminative capabilities. Statistical significance tests, such as paired t-tests, will be conducted to validate performance improvements.

Experiment Setup and Implementation

The higher overview of steps involved in classifying MI tasks is shown below.

Proposed Approach

Results and Analysis

Results have not yet been finalized.

Conclusion

This proposed research approach focuses on classifying wrist Motor Imagery(MI) tasks using EEG signals, specifically pronation and supination movements. The approach consists of preprocessing, feature extraction, model training, and evaluation to enhance classification accuracy. This study aims to improve MI-based BCI systems by addressing challenges in classifying single-joint movements, contributing to advancements in assistive technologies and neurorehabilitation.

Publications

This project has not yet been published.