Design and FPGA Implementation of an Efficient Architecture for Noise Removal in ECG Signals Using Lifting-Based Wavelet Denoising

Abstract

Noise removal is the most crucial pre-processing step for present-generation biomedical wearable electrocardiogram (ECG) patches and devices to provide efficient detection and monitoring of cardiac arrhythmias. This paper proposes a hardware-efficient and multiplier-less FPGA-based ECG noise removal architecture based on lifting-based wavelet denoising that employs a universal threshold level-dependent function in combination with soft thresholding to produce a noise-free ECG signal. The paper also proposes a modified lifting-based discrete wavelet transform (DWT) algorithm that is multiplier-less and provides a one-step equation for the calculation of the forward output coefficients and the inverse output coefficients. Since a comparator circuit is a very complicated circuitry in VLSI implementation, an optimized median calculation and soft thresholding block with no compare operations for wavelet-based thresholding is proposed. The ECG data is collected from the MIT-BIH arrhythmia database and the ECG noises from the MIT-BIH noise stress database. The proposed denoising technique for the ECG signal is tested on MATLAB which achieves an average improvement in SNR of 7.4 dB and an MSE of 0.0206. The FPGA implementation is performed on the Nexys 4 DDR board, and the proposed wavelet-based denoising architecture results in lower hardware utilization and a relatively high operating frequency of 166 MHz when compared to existing ECG denoising architectures.