EVOLUTIONARY BASED ATTENTION NESTED U NET ARCHITECTURE FOR ECG -BASED HEARTBEAT CLASSIFICATION

Abstract

Electrocardiogram (ECG) signals are widely used for diagnosing cardiac disorders, yet their reliability is often affected by various noise sources that distort waveform morphology and reduce diagnostic accuracy. Accurate heartbeat classification therefore remains a critical challenge in automated cardiac monitoring systems. This study aims to develop a robust and efficient ECG classification framework capable of enhancing noise resilience and improving diagnostic precision. The proposed method integrates an Improved Empirical Mode Decomposition (IEMD) technique for effective noise reduction, followed by Fast Fourier Transform (FFT) for extracting frequency-domain features. IEMD produces cleaner Intrinsic Mode Functions (IMFs) compared to conventional EMD, leading to better signal decomposition, while FFT provides a compact and discriminative representation of ECG characteristics. For classification, an Evolutionary Based Attention Nested U-Net (EBANU-Net) is employed, where the U-Net architecture is enhanced.