Denoising Convoluted Neural Network-Assisted ECG Signal Watermarking for Secure Transmission in E-Health-Care Applications

Jyoti Rani, Ashima Anand, Shivendra Shivani

Abstract: Electrocardiogram (ECG) signals are vulnerable to tampering, forgery, and unauthorized access due to their digital transmission and lack of built-in security. Digital watermarking helps protect ECG data by embedding hidden information for authentication and tamper detection without affecting its clinical quality. This study presents a novel reversible watermarking architecture utilizing ECG signals for secure data transfer of telemedicine, tackling issues, such as inadequate robustness and signal deterioration found in current methodologies. Pan–Tompkins++ transforms the 1-D ECG into 2-D extracted features, redundant discrete wavelet transform, and multiresolution singular value decomposition, and encrypted watermarks are hidden in the transformed components. Denoising convoluted neural network denoises the retrieved watermarks, offering robustness against common attacks while preserving signal quality. Experimental findings indicate a notable improvement, with a peak signal-to-noise ratio n value of 83.48 dB and normalized correlation score of 0.9991, surpassing current methodologies, with enhanced resilience to prevalent attacks, including Gaussian noise, speckle noise, and JPEG compression.

External IDs:dblp:journals/ieeemm/RaniAS25