IRO Journals
Journal of Soft Computing Paradigm

Enhancing ECG Classification Using GAN-Augmented DCT Features and Attention Mechanisms

Open Access
Published: 16 July, 2026
Pages: 266-279
  • Vharshni K. , Vharshni K.
    Department of Computer Science and Engineering, College of Engineering, Guindy, Anna University, Chennai, India
    Department of Computer Science and Engineering, College of Engineering, Guindy, Anna University, Chennai, India
  • Angelin Gladston Angelin Gladston
    Department of Computer Science and Engineering, College of Engineering, Guindy, Anna University, Chennai, India
    Department of Computer Science and Engineering, College of Engineering, Guindy, Anna University, Chennai, India
view PDF
view PDF

How to Cite

K., Vharshni, and Angelin Gladston. 2026. “Enhancing ECG Classification Using GAN-Augmented DCT Features and Attention Mechanisms”. Journal of Soft Computing Paradigm 8 (3): 266-79. https://doi.org/10.36548/jscp.2026.3.006.
ACM ACS APA ABNT Chicago Harvard IEEE MLA Turabian Vancouver AMA

Download Citation

Endnote/Zotero/Mendeley (RIS) BibTeX

Keywords

Electrocardiogram (ECG)
Generative Adversarial Network (GAN)
Discrete Cosine Transform (DCT)
Self-Attention Mechanism
Deep Learning
ECG Classification

Abstract

Electrocardiogram (ECG) classification plays a major role in the early detection and diagnosis of heart disorders. The present ECG signals are noisy, temporally complex and also inter-class similar in characteristics, this makes the classification very complex. In this research, a deep learning approach is proposed for ECG classification with the integration of an attention-enhanced convolutional neural network model and a DCT-based feature extractor. Firstly, ECG5000 dataset is pre-processed through Z-score normalization, after which the dataset is converted to the DCT space in order to obtain a compact spectral feature and at the same time reduce noise while preserving discriminative spectral feature components. Next, the obtained features are normalized using spectral normalization and inputted into the proposed deep learning approach for classification. Th proposed model employs a patch-based self-attention mechanism focusing on specific spectral regions for diagnosis followed by convolutional feature extraction and global average pooling. The accuracy, precision, recall, F1-score, AUROC and AUPRC metrics are used to evaluate the proposed model on ECG5000 dataset. The experimental results show a classification accuracy of 95.6%, demonstrating that DCT-based spectral features and attention-guided deep learning collectively provide powerful, fast, and reliable tools for ECG automatic classification and cardiac abnormality detection.

References

  1. Feng, Haotian, Qiang Shen, Rui Song, Lida Shi, and Hao Xu. "ATFA: Adversarial Time–Frequency Attention Network for Sensor-Based Multimodal Human Activity Recognition." Expert Systems with Applications 2024, vol 236: 121296.
  2. Ma, Xianping, Xiaokang Zhang, Man-On Pun, and Bo Huang. "A Unified Framework with Multimodal Fine-Tuning for Remote Sensing Semantic Segmentation." IEEE Transactions on Geoscience and Remote Sensing 2025: 1-15.
  3. Aydoğdu, Özge, and Murat Ekinci. "An Approach for Streaming Data Feature Extraction Based on Discrete Cosine Transform and Particle Swarm Optimization." Symmetry 2020, vol 12, no. 2: 299.
  4. Tang, Peihao, Zhen Li, Xuanlin Wang, Xueping Liu, and Peng Mou. "Time Series Data Augmentation for Energy Consumption Data Based on Improved Timegan." Sensors 2025, vol 25, no. 2: 493.
  5. Haboub, Amine, Hamza Baali, and Abdesselam Bouzerdoum. "DCT-Based Channel Attention for Multivariate Time Series Classification." IEEE Open Journal of the Computer Society 2025, vol 6: 1110-1120.
  6. Mi, Zhongjie, Xinghao Jiang, Tanfeng Sun, and Ke Xu. "GAN-Generated Image Detection with Self-Attention Mechanism Against GAN Generator Defect." IEEE Journal of Selected Topics in Signal Processing 2020, vol 14, no. 5: 969-981.
  7. Hossain, Khondker Fariha, Sharif Amit Kamran, Alireza Tavakkoli, Lei Pan, Xingjun Ma, Sutharshan Rajasegarar, and Chandan Karmaker. "ECG-Adv-GAN: Detecting ECG Adversarial Examples with Conditional Generative Adversarial Networks." In 2021 20th IEEE International Conference on Machine Learning and Applications (ICMLA), IEEE, 2021: 50-56.
  8. Nankani, Deepankar, and Rashmi Dutta Baruah. "Improved Diagnostic Performance of Arrhythmia Classification Using Conditional Gan Augmented Heartbeats." In Generative Adversarial Learning: Architectures and Applications, Cham: Springer International Publishing, 2022: 275-304.
  9. Wyawahare, Medha, Milind Rane, Tushar Bhandari, Pratham Bisen, and Sandesh Bellale. "DFD-TL-GAN: Deepfake Detection Using Transfer Learning and GAN-Augmented Datasets." In 2025 2nd Asian Conference on Intelligent Technologies (ACOIT), IEEE, 2025, 1-7.
  10. Khade, Smita, Shilpa Gite, and Biswajeet Pradhan. "Robust Iris Liveness Detection Against Known and Adversarial Attacks Using GAN and Efficient Net." In Advancements in IoT Sensors and Security: Harnessing AI/ML for Secured IoT Sensor Data, Cham: Springer Nature Switzerland, 2026: 145-171.
  11. Dau, H. A., Bagnall, A., Kamgar, K., Yeh, C. C. M., Zhu, Y., Gharghabi, S., et al., "The UCR Time Series Classification Archive," October 2018. Available
  12. https://www.kaggle.com/datasets/salsabilahmid/ecg50000/data