ECG Compressed Measurements and Fast Reconstruction with Sparse Sensing Matrix and Lightweight Neural Network

Real-time healthcare monitoring is critical for wearable and IoT-based medical devices, particularly in the context of ECG signal processing. In this work, we explore the application of compressive sensing (CS) for efficient ECG signal acquisition and propose a novel structured binary block sparse sensing matrix for compressed ECG measurements. This structured approach ensures that key ECG features are preserved even at high compression levels. For signal reconstruction, we introduce a lightweight neural network (NN)-based decoder that guarantees accurate recovery with significantly reduced computational complexity. Experimental results demonstrate that our method achieves an 5.95 compression ratio while maintaining a PRD of less than 9%, ensuring high reconstruction quality and minimal signal distortion. Moreover, our NN-based approach reconstructs a single ECG segment in an average time of 82.6 μ s, which is approximately 2914 times faster than traditional iterative methods like Basis Pursuit Denoising (BPDN), making it ideal for real-time applications. By balancing computational efficiency and signal fidelity, our approach offers a practical and scalable solution for wearable ECG monitoring in resource-limited environments. This work highlights the potential of structured sparse sensing and deep learning-based reconstruction for medical measurement systems, paving the way for efficient, accurate, and low-power ECG monitoring in real-world healthcare applications.