<![CDATA[Ambulatory Holter electrocardiography (ECG) enables continuous monitoring for detecting transient arrhythmias; however, its diagnostic reliability is significantly degraded by motion artifacts and electromyographic (EMG) interference. Under severe motion artifact conditions, prior studies report that ambulatory ECG SNR can fall below −10 dB , although SNR levels vary substantially depending on activity type and electrode placement, reducing usable data segments and impairing arrhythmia detection. While advanced denoising methods such as wavelet transforms and deep learning achieve high accuracy, their computational complexity limits real-time deployment in resource-constrained embedded systems. This reveals a critical gap in lightweight methods that jointly optimize noise suppression, morphological preservation, and downstream diagnostic performance. This study proposes a computationally efficient IIR Butterworth bandpass filtering framework for real-time IoT-based Holter ECG systems. The system combines three-lead ECG acquisition, embedded processing on an ESP32, and real-time visualization. Performance is assessed using SNR, mean squared error (MSE), Pearson correlation, and confusion matrix-based detection metrics on ten male participants under controlled motion and muscle artifact conditions. Results demonstrate statistically significant SNR improvements for motion artifacts (ΔSNR = 9.47 ± 1.96 dB, t(9) = 15.28, p < 0.001) and EMG artifacts (ΔSNR = 16.73 ± 0.91 dB, t(9) = 58.11, p < 0.0001). Post-filtering morphological fidelity was high, with mean Pearson correlation of 0.963 for motion artifacts and 0.945 for muscle artifacts. These signal quality improvements translated into 95.3% post-filtering arrhythmia detection accuracy (sensitivity: ≈96.0%, specificity: ≥97.0%, F1-score: ≥95.0%), significantly exceeding the 70% minimum performance threshold adopted in this study as a conservative screening criterion (t(9) = 29.7, p < 0.001). Despite dataset limitations (n = 10), the proposed framework provides an effective trade-off between computational efficiency and diagnostic reliability, supporting scalable and real-time ambulatory ECG monitoring for early arrhythmia screening.]]>
