Advanced Signal Processing Algorithm-Based State Estimation and Denoising of ECG Signals for Biomedical Applications

State estimation of electrocardiography (ECG) signals plays a crucial role in effective cardiac monitoring systems and assists in the early detection of abnormalities. However, ECG signals are often contaminated by interfering noise such as baseline wander, power-line interference, and motion artifacts, which significantly hinder precise state estimation. To address these challenges, this paper proposes a particle filter (PF)-based method for ECG signal estimation, which effectively preserves clinically relevant waveform characteristics under nonlinear and non-stationary noise conditions. The technique is implemented in MATLAB using a synthetic ECG signal and is evaluated using fast Fourier transform, spectrogram analysis, power spectral density, autocorrelation, and wavelet transform. A comparative analysis reveals that the proposed PF-based method significantly outperforms baseline denoising techniques. Quantitatively, the process achieves a mean square error (MSE) of 0.001382, representing a 48.5% improvement over the variant without the attention mechanism (MSE =0.002682). Additionally, the signal-to-noise ratio is enhanced by 6.3 dB, reaching 15.87 dB, the highest among all tested methods. The percentage root mean square difference is also reduced to 4.62%, indicating low signal distortion. Furthermore, the accuracy of R-peak identification reaches 98.5%, ensuring a better extraction of cardiac features. The proposed method demonstrates a runtime of 0.75 s for a 10-s ECG segment, confirming its suitability for real-time biomedical applications. The simulation results validate that the PF-based state estimation framework is not only more precise and robust to noise but also computationally efficient, making it ideal for ECG signal processing in wearable health monitoring and real-time cardiac diagnostics.