Optimizing Subject Independent Classification Performance in class imbalanced, EMG-Based Gesture Recognition
The primary goal is to enhance the classification accuracy of EMG signals for static hand gestures, specifically focusing on Class 1 to Class 7. We'll employ advanced neural network architectures and machine learning techniques for this. Class 0 can be used as a baseline or as deemed appropriate.
Improving the accuracy of EMG-based gesture recognition has significant implications for BCI applications, including assistive technologies and human-computer interaction.
The dataset contains raw EMG data from 36 subjects performing static hand gestures(Open data). Each subject executed two series of 6 or 7 basic gestures. Each gesture lasted for 3 seconds, with a 3-second pause between gestures. Data was collected using a MYO Thalmic bracelet equipped with eight sensors.
-
Columns:
- Time - Time in ms 2-9) Channel - Eight EMG channels from MYO Thalmic bracelet
- Class - Gesture labels:
- 0: Unmarked data
- 1: Hand at rest
- 2: Hand clenched in a fist
- 3: Wrist flexion
- 4: Wrist extension
- 5: Radial deviations
- 6: Ulnar deviations
- 7: Extended palm (not performed by all subjects)
-
Additional Column:
- Label: Refers to the subject who performed the experiment
You can download the .csv files for the dataset from this Google Drive link.
conda 23.3.0, Python 3.11.4, torch 2.0.1+cu117
or use jyk.yaml the copy of my own env, and make conda env.