Associate Professor, Department of Electronics, Baselios Poulose II Catholicose College, Piravom, Kerala, India
Email: babypaul@bpccollege.ac.in
This study introduces a novel adaptive filtering technique to effectively reduce noise in electrocardiogram (ECG) signals by leveraging a Genetic Algorithm (GA)-tuned Signed Data-Least Mean Square (SD-LMS) algorithm. ECG signals are essential in the diagnosis and monitoring of cardiovascular conditions, including arrhythmias, ischemic events, and other heart abnormalities. However, their accuracy is often compromised by noise sources, primarily baseline wander (BLW) and power line interference (PLI), which can obscure critical features such as the P, QRS, and T waveforms that are crucial for clinical interpretation. BLW, a low-frequency noise, is commonly introduced by patient movement and respiration, while PLI, a high-frequency interference often at 50 or 60 Hz, originates from nearby electrical devices or improper grounding. These noises complicate ECG analysis and can lead to misinterpretation or missed diagnoses if not adequately filtered.
Existing noise removal techniques, while useful, often lack adaptability and can be computationally demanding, making them unsuitable for real-time applications in embedded systems and portable medical devices. Adaptive filters, such as those in the Least Mean Square (LMS) algorithm family, are widely regarded for their ability to dynamically adjust filter parameters in response to noise changes. The SD-LMS variant, in particular, is computationally efficient, making it well-suited for resource-limited applications. However, a limitation of LMS-based filters is that their performance depends on the step size parameter (ยต), which governs the speed and accuracy of convergence. An inappropriate step size can lead to slow convergence or even filter divergence, negatively impacting the quality of noise reduction.
To address this limitation, the proposed method uses a Genetic Algorithm (GA) to optimize the step size of the SD-LMS filter. GAs are heuristic optimization techniques inspired by natural selection and are effective for complex search and optimization tasks. In this approach, the GA iteratively tunes the step size based on its ability to maximize the Signal-to-Noise Ratio (SNR), resulting in a filter that adapts efficiently to noise without requiring manual parameter adjustments. This optimization process enables the GA-tuned SD-LMS filter to achieve robust noise reduction across various noise profiles.
The proposed filtering method was validated using records from the MIT-BIH Arrhythmia and Noise Stress Test databases, which are standard datasets in biomedical research. The GA-tuned SD-LMS filter demonstrated an average SNR improvement of 10.754 dB for BLW and 24.08 dB for PLI, significantly surpassing traditional filtering techniques. Moreover, the high correlation coefficients achieved indicate that this filter preserves essential ECG features, enhancing the reliability of ECG signal analysis in clinical and real-time monitoring applications. This adaptive filtering approach shows promise for integration into wearable ECG devices, where maintaining signal fidelity in noisy environments is critical for patient care.
Keywords: ECG signal processing, Adaptive filtering, Genetic Algorithm, SD-LMS, Baseline wander, Power line interference, Signal-to-Noise Ratio, MIT-BIH database.
How to cite this article: Paul B. Enhanced ECG Signal Processing Through GA-Optimized SD-LMS Adaptive Filtering. Int J Drug Deliv Technol. 2026;16(7s): 654-661; DOI: 10.25258/ijddt.16.7s.69
Source of support: None
Conflict of interest: None