ECG Signal Analysis for Arrhythmia Recognition Using Wavelet Scattering Features and Linearized Attention Text (Wavelet-Linformer)
DOI:
https://doi.org/10.29284/d67et920Keywords:
Signal processing, Arrhythmia classification, Linformer, ECG, Biomedical Informatics, Deep learning, Wavelet Scattering Transform.Abstract
Cardiac arrhythmias contribute vastly to cardiovascular mortality and morbidity, and automated electrocardiogram (ECG) analysis is essential to support timely detection in high-volume monitoring environments. Deep learning models have achieved strong performance for arrhythmia classification, but end-to-end convolutional or Transformer architectures are often computationally expensive, data-hungry, and difficult to interpret. This work proposes a Wavelet-Linformer, a hybrid model that combines the Wavelet Scattering Transform (WST) with a linearized self-attention encoder for beat-level arrhythmia classification. WST provides fixed, multiscale time–frequency features with built-in stability to time shifts, reducing the need to learn low-level filters from raw signals. These features are fed to a Linformer encoder, approximating self-attention through low-rank projections of the key and value matrices to reduce the quadratic to linear cmplexity in sequence length and enabling more efficient inference. We evaluate Wavelet-Linformer on the MIT–BIH Database and AAMI EC57 five-class grouping under an inter-patient protocol, and compare it with three strong baselines: ResNet-1D, LSTM with attention, and a pure Transformer encoder. Experiments show that Wavelet-Linformer achieves slight but consistent improvements in accuracy, macro F1, and macro AUROC over all baselines, while using fewer parameters than the full Transformer. Analysis of rare classes (fusion and unknown/artefact beats) indicates that these gains extend to under-represented arrhythmia types. Attention maps over WST features highlight time–frequency regions aligned with characteristic ECG components, suggesting that the model’s decisions are driven by physiologically meaningful patterns and supporting its potential for interpretable and resource-aware arrhythmia screening.
References
1. World Health Organization. Cardiovascular diseases (CVDs) [Internet]. Geneva: World Health Organization; 2025 Jul 31 [cited 2025 Dec 7]. Available from: https://www.who.int/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds)
2. Srinivasan NT, Schilling RJ. Sudden cardiac death and arrhythmias. Arrhythm Electrophysiol Rev. 2018;7(2):111–7.
3. Moody GB, Mark RG. The impact of the MIT-BIH Arrhythmia Database. IEEE Eng Med Biol Mag. 2001;20(3):45–50.
4. Association for the Advancement of Medical Instrumentation. ANSI/AAMI EC57:2012/(R)2020: Testing and reporting performance results of cardiac rhythm and ST-segment measurement algorithms. Arlington (VA): AAMI; 2020.
5. Ansari Y, Mourad O, Qaraqe K, Serpedin E. Deep learning for ECG arrhythmia detection and classification: an overview of progress for period 2017–2023. Front Physiol. 2023;14:1246746.
6. Boulif A, Ouladsine M, Boulif I, Bensid A. A literature review: ECG-based models for arrhythmia diagnosis using artificial intelligence techniques. Bioinform Biol Insights. 2023;17:11779322221149600.
7. Xiao Q, Zhang Q, Wang H, Yang C, Wang S. Deep learning-based ECG arrhythmia classification: a survey of existing methods. Appl Sci. 2023;13(8):4964.
8. Liu Z, Yao G, Zhang Q, Zhang J, Zeng X. Wavelet scattering transform for ECG beat classification. Comput Math Methods Med. 2020;2020:3215681.
9. Mallat S. Group invariant scattering. Commun Pure Appl Math. 2012;65(10):1331–98.
10. Bruna J, Mallat S. Invariant scattering convolution networks. IEEE Trans Pattern Anal Mach Intell. 2013;35(8):1872–86.
11. Dhyani A, Dwivedi RK, Basu A, Kumar A. Review of ECG arrhythmia detection techniques based on wavelet transform and time–frequency analysis. Chaos Solitons Fractals X. 2023;8:100115.
12. Mohonta SK, Guha R, Dhar PK. Electrocardiogram-based arrhythmia classification using phase-related time–frequency representations with 2D convolutional neural network. Biomed Signal Process Control. 2022;71:103107.
13. Acharya UR, Oh SL, Hagiwara Y, Tan JH, Adam M, Gertych A, et al. A deep convolutional neural network model to classify heartbeats. Comput Biol Med. 2017;89:389–96.
14. Wu M, Lu Y, Yang W, Wong SY. A study on arrhythmia via ECG signal classification using the convolutional neural network. Front Comput Neurosci. 2021;14:564015.
15. Yildirim Ö. A novel wavelet sequence based on deep bidirectional LSTM network model for ECG signal classification. Comput Biol Med. 2018;96:189–202.
16. Saadatnejad S, Oveisi M, Hashemi M. LSTM-based ECG classification for continuous monitoring on personal wearable devices. IEEE J Biomed Health Inform. 2019;24(2):515–23.
17. Oh SL, Ng EYK, Tan RS, Acharya UR. Automated diagnosis of arrhythmia using combination of CNN and LSTM techniques with variable length heartbeats. Comput Biol Med. 2018;102:278–87.
18. Alamatsaz MH, Pirzadeh H, Asl BM. A lightweight hybrid CNN–LSTM explainable model for automatic classification of ECG arrhythmias. Biomed Signal Process Control. 2024;88:105555.
19. Tay Y, Dehghani M, Bahri D, Metzler D. Efficient Transformers: a survey. ACM Comput Surv. 2022;55(6):109:1–28.
20. Wang S, Li BZ, Khabsa M, Fang H, Ma H. Linformer: self-attention with linear complexity. arXiv Preprint. 2020;arXiv:2006.04768.
21. Arabi Z, Pourbahrami S, Abedi A. Hybrid CNN-transformer architecture with adaptive positional embedding for interpretable ECG arrhythmia classification. SSRN Preprint. 2025. doi:10.2139/ssrn.5245392.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
This work is licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
