Lightweight GA-Optimized H-HEDPAF Algorithm with mRMR for Robust ECG Detection in Secure IoT Medical Sensing

Conventional electrocardiogram (ECG) detection methods encounter several challenges in IoT-based monitoring: noise sensitivity, lack of flexibility, and high computational cost. To tackle these issues, this paper presents a Hybrid Hilbert Envelope–Derivative Phase with Autocorrelation-Guided Adaptive Fusion (H-HEDPAF) algorithm featuring an adaptive sequential fusion strategy. This algorithm is optimized with a lightweight genetic algorithm (GA) that enables automatic adjustment of the contribution of each individual GA chromosome to each fusion parameter. We also conduct an ablation study to evaluate the confidence contribution of each detector’s weight. Moreover, Minimum Redundancy Maximum Relevance (mRMR) feature selection reduces computational overhead. Furthermore, the paper proposes an IoT-based remote medical monitoring system that incorporates the H-HEDPAF algorithm alongside biomedical and environmental sensors, utilizing Blynk cloud and HTTPS protocols to ensure secure real-time monitoring. The suggested H-HEDPAF algorithm was validated on 40 patients total: 20 records from the MIT-BIH Arrhythmia Database and 20 patients from the proposed IoT-based monitoring system. On the MIT-BIH database, the algorithm achieved 97.55% accuracy, 98.40% precision, 99.87% sensitivity, and 98.51% F1-score. On the proposed IoT system, which collected data from 20 patients, it achieved 90% accuracy, 77.78% specificity, 84.61% sensitivity, and 91.66% F1-score. Under realistic noise (SNR = -6 dB) from the MIT-BIH Noise Stress Test Database, sensitivity remained > 98% and F1-score > 97%. The mean absolute error for RR interval estimation on MIT-BIH data was 0.285 ms. These results demonstrate robustness in noisy environments and real-time IoT deployment. Clinical validation on larger cohorts remains future work.