Skip to content

Latest commit

 

History

History
64 lines (46 loc) · 2.92 KB

README.md

File metadata and controls

64 lines (46 loc) · 2.92 KB
Raw

EA-wLTL (Work in Progress)

update 202406 life gets wild, cant code no more, ive tried.

Experimenting combining Euclidean Alignment (EA) and weighted LTL to classify MI-based EEG

All results shown are still being developed

Non-EA and EA EEG trials of one trial

Updated on 25th March, 2021:

  • Preprocessed data
  • Comparing 6 approaches

This is a comparison of non-aligned (black) vs aligned (red) EEG trials from a single trial of all electrodes

Visualization of non-alinged features vs aligned features using t-sne

SNE_plot Not the expected result, the result for EA for target subject (red dot) is expected to more scattered

Effect of EA on LDA and SVM

This section compares effects of doing EA (Euclidean Alignment) using LDA and SVM as classifier, each subject alternately acts as target while the other 8 act as source when EA is applied.

Objective

  1. Despite the model, and whether or not EA is applied, using same number of trials won't improve the result
  2. Using non-EA source trials to train target will worsen accuracy
  3. using EA source trials to train target will improve accuracy

Evaluation scheme

evaluation_scheme

Result

evaluation_scheme

Conclusion:

  1. Objective 1 is proofed by comparing pattern 1 and 2, the difference between the two is negligible, on either classifier.
  2. Objective 2 and 3 can be observe by comparing pattern 3 and pattern 4

Comparing 6 Approaches

Comparison of six different approaches they are:

  1. CSP-SVM
  2. CSP-LDA
  3. EA-CSP-LDA
  4. EA-CSP-SVM
  5. CSP-wLTL
  6. EA-CSP-wLTL Here wLTL stands for Weighted Logistic Transfer Learning[3]

10_20_barplot 30_40_barplot

Classification of left and right hand imagery task. Different number of target training trials from 10 trials (5 each class) to 40 trials are used to observe the effect it has on accuracy on different approaches.

Number of Source Data vs Accuracy

lineplot

One significant result happened on subject 8 where wLTL perform better than the rest of other approaches, this agrees with the study on [3] that wLTL is more pronounced on subject with poor performance.

Reference

  1. He, H., & Wu, D. (2020). Transfer Learning for Brain-Computer Interfaces: A Euclidean Space Data Alignment Approach. IEEE Transactions on Biomedical Engineering, 67(2), 399–410. https://doi.org/10.1109/TBME.2019.2913914
  2. Wu, D., Peng, R., Huang, J., & Zeng, Z. (2020). Transfer Learning for Brain-Computer Interfaces: A Complete Pipeline. 1–9. http://arxiv.org/abs/2007.03746
  3. Azab, A. M., Mihaylova, L., Ang, K. K., & Arvaneh, M. (2019). Weighted Transfer Learning for Improving Motor Imagery-Based Brain-Computer Interface. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 27(7), 1352–1359. https://doi.org/10.1109/TNSRE.2019.2923315