Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Teager-Kaiser Energy Methods For EEG Feature Extraction In Biomedical Applications

Ioanna Chourdaki, Kleanthis Avramidis, Christos Garoufis, Athanasia Zlatintsi, Petros Maragos

TL;DR

This work investigates the non-linear Teager-Kaiser Energy Operator (TKEO) for modeling the underlying energy dynamics of EEG in three representative tasks: motor imagery, emotion recognition, and epilepsy detection, and suggests that combining TKEO features with conventional ones improves Balanced Accuracy by approximately 15 percent in epilepsy detection.

Abstract

Electroencephalography (EEG) signals are inherently non-linear, non-stationary, and vulnerable to noise sources, making the extraction of discriminative features a long-standing challenge. In this work, we investigate the non-linear Teager-Kaiser Energy Operator (TKEO) for modeling the underlying energy dynamics of EEG in three representative tasks: motor imagery, emotion recognition, and epilepsy detection. To accommodate the narrowband nature of the operator, we employ Gabor filterbanks to isolate canonical frequency bands, followed by the Energy Separation Algorithm to decompose the TKEO output into amplitude envelope and instantaneous frequency components. We then derive a set of energy descriptors based on this demodulation and compare their classification performance against established EEG features. The proposed TKEO-based pipeline offers an intuitive, physiologically grounded framework for capturing EEG signal dynamics, while remaining simple, training-free, and data-efficient. Our findings suggest that combining TKEO features with conventional ones improves Balanced Accuracy by approximately 15 percent in epilepsy detection, yields modest gains in motor imagery, and achieves on par performance in emotion recognition, reflecting the pipeline's ability to capture transient neural dynamics.

Teager-Kaiser Energy Methods For EEG Feature Extraction In Biomedical Applications

TL;DR

This work investigates the non-linear Teager-Kaiser Energy Operator (TKEO) for modeling the underlying energy dynamics of EEG in three representative tasks: motor imagery, emotion recognition, and epilepsy detection, and suggests that combining TKEO features with conventional ones improves Balanced Accuracy by approximately 15 percent in epilepsy detection.

Abstract

Electroencephalography (EEG) signals are inherently non-linear, non-stationary, and vulnerable to noise sources, making the extraction of discriminative features a long-standing challenge. In this work, we investigate the non-linear Teager-Kaiser Energy Operator (TKEO) for modeling the underlying energy dynamics of EEG in three representative tasks: motor imagery, emotion recognition, and epilepsy detection. To accommodate the narrowband nature of the operator, we employ Gabor filterbanks to isolate canonical frequency bands, followed by the Energy Separation Algorithm to decompose the TKEO output into amplitude envelope and instantaneous frequency components. We then derive a set of energy descriptors based on this demodulation and compare their classification performance against established EEG features. The proposed TKEO-based pipeline offers an intuitive, physiologically grounded framework for capturing EEG signal dynamics, while remaining simple, training-free, and data-efficient. Our findings suggest that combining TKEO features with conventional ones improves Balanced Accuracy by approximately 15 percent in epilepsy detection, yields modest gains in motor imagery, and achieves on par performance in emotion recognition, reflecting the pipeline's ability to capture transient neural dynamics.

Paper Structure

This paper contains 15 sections, 5 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Teager-Kaiser Energy Operator
  4. Filterbank Decomposition
  5. Energy Separation Algorithm
  6. TKEO Feature Extraction
  7. Experimental Setup
  8. Datasets
  9. Preprocessing
  10. Baseline Feature Extraction
  11. Evaluation Protocol
  12. Results & Discussion
  13. Main Results
  14. Additional Experiments
  15. Conclusion

Figures (3)

  • Figure 1: (a) CP4 channel signal from the BCI-IV 2a dataset in the Alpha band. (b) Square root of TKEO. (c) Estimated amplitude envelope using DESA-1. (d) Estimated instantaneous frequency using DESA-1, expressed as a fraction of $\pi$.
  • Figure 2: Performance of different filterbanks on BCI-IV 2a.
  • Figure 3: Absolute SHAP values of frequency bands indicating their contributions to fused-band TKEO predictions across EEG channels in BCI-IV 2a; higher SHAP values represent greater contribution.