Related Experiment Video
Updated: Oct 19, 2025
05:48
Author Spotlight: Investigating the Impact of Emotional Prosodies on Voice Recognition and Perception
Published on: August 9, 2024
1.7K
PrimePatNet87: Prime pattern and tunable q-factor wavelet transform techniques for automated accurate EEG emotion
Abdullah Dogan1, Merve Akay2, Prabal Datta Barua3
1Department of Computer Engineering, Middle East Technical University, Ankara, Turkey.
Computers in Biology and Medicine
|September 20, 2021
Summary
A novel hand-crafted network, PrimePatNet87, efficiently classifies emotions from electroencephalogram (EEG) signals. This model achieves over 99% accuracy, overcoming limitations of computationally intensive deep learning methods for emotion recognition.
Area of Science:
- Neuroscience and Artificial Intelligence
- Signal Processing and Machine Learning
Background:
- Deep learning models for electroencephalogram (EEG) based emotion recognition are computationally intensive and face classification performance challenges.
- Existing methods often require extensive training time and struggle to achieve high accuracy in emotion classification tasks.
Purpose of the Study:
- To develop a computationally efficient, hand-crafted cognitive model for accurate emotion classification using EEG signals.
- To overcome the limitations of deep learning models in terms of training time and classification performance.
Main Methods:
- Utilized tunable q-factor wavelet transform (TQWT) for sub-band decomposition of EEG signals.
- Employed a novel prime pattern and statistical feature generator, followed by minimum redundancy maximum relevance (mRMR) feature selection and Support Vector Machine (SVM) classification.
- Developed the PrimePatNet87 model, incorporating feature extraction, selection, and iterative majority voting for robust emotion classification.
Main Results:
- The PrimePatNet87 model achieved over 99% classification accuracy across DEAP, DREAMER, and GAMEEMO datasets.
- Demonstrated high performance using leave-one-subject-out (LOSO) validation, indicating strong generalization capabilities.
- The proposed network effectively extracts and selects relevant features for precise emotion recognition.
Conclusions:
- The developed PrimePatNet87 model offers an accurate and efficient alternative to deep learning for EEG-based emotion classification.
- The novel combination of prime patterns and TQWT provides a robust framework for real-world emotion recognition applications.
- The hand-crafted approach proves effective in achieving high classification accuracy while reducing computational load.
