METHODS OF CARDIAC SIGNAL RESTORATION AND COMPLEX PROCESSING

Accurate restoration and complex processing of cardiac signals, i.e., electrocardiograms (ECG), are extremely crucial for reliable diagnosis and automatic detection of cardiovascular diseases. The present work gives a complete framework that includes signal restoration, denoising, feature extraction, and classification to enhance the precision of ECG analysis. Reconstruction of missing and corrupted cardiac segments is accomplished by Hermite polynomial interpolation with Chebyshev nodes and adaptive filtering by the Least Mean Squares (LMS) algorithm. Denoising is carried out by hybrid wavelet–CEEMDAN decomposition for attenuation of baseline drift, motion artifacts, and high-frequency noise without compromising morphological features. The proposed method then continues with the extraction of effective temporal, spectral, and nonlinear features—RR intervals, power spectral density, and sample entropy—prior to dimensionality reduction via Principal Component Analysis (PCA). The classification is finally carried out using deep convolutional neural networks (CNNs) with the MITBIH Arrhythmia Database as the training data. Experimental validation illustrates significant improvement in signal-to-noise ratio (SNR), percent root-mean-square difference (PRD), and F1- score compared to conventional techniques. The results validate that the proposed combination of advanced signal restoration and deep learning structures is a viable solution to precise ECG analysis and real-time cardiac monitoring.