Data-driven compressive sensing approach for ECG signals in IoT healthcare applications

Abstract

The rapid adoption of Internet of Things (IoT) technologies in healthcare has transformed patient monitoring, particularly in continuous ECG monitoring for early detection of cardiac abnormalities. However, traditional ECG monitoring methods face challenges such as high data volume, power consumption, and transmission inefficiencies, complicating real-time monitoring in resource-constrained environments. This study introduces a novel data-driven compressive sensing framework designed for ECG signal processing in IoT healthcare applications. The framework incorporates a Data-Driven Sensing Matrix (DSM) and Binary Thresholding Matrix (BTM) to optimize hardware efficiency while maintaining high reconstruction accuracy. DSM leverages machine learning to adapt to ECG signal properties, while BTM employs a novel thresholding technique for efficient hardware implementation. Additionally, overcomplete dictionaries, such as Gaussian and K-SVD, enhance sparsity and reconstruction accuracy. Performance validation using the MIT-BIH Arrhythmia Database demonstrates that the reconstructed signal preserves key features, with Percent Root Mean Square Difference values below 9% at compression ratios up to 85%. Comparative evaluations confirm the superiority of DSM and BTM over conventional sensing matrices like Random Gaussian, Bernoulli Binary, and Signed Matrices in compression efficiency and reconstruction accuracy. These findings highlight the potential of data-adaptive compressive sensing for energy-efficient, secure, and real-time ECG monitoring in IoT-driven healthcare. The proposed BTM, with its low computational requirements and efficient hardware integration, addresses key challenges in wearable and portable ECG devices, ensuring scalable and reliable performance in real-world applications.