Hardware Requirements of Different Wavelet Filter Bank Architectures for ECG Signal Denoising

Electrocardiogram (ECG) is a non-invasive and extensively used method to analyze cardiac health status. Cardiovascular diseases (CVDs) severely affect cardiac health and become the leading cause of death. Accurate and timely detection of ECG irregularities can avoid the severity of CVDs. Detection of CVDs using an ECG signal is tricky as noises in an ECG signal can mislead the diagnostic. Wavelet-transform-based ECG denoising techniques are becoming more popular due to low computational complexity. This paper compares the computational complexity of wavelet filter banks such as binomial quadrature mirror filter, lattice filter, lifting ladder filter, Mallat filter, modified Mallat structure, short-length filter, and transversal filter while denoising an ECG signal. The complexity of wavelet filter banks is computed based on the number of adders, multipliers, and delay elements required to realize the architecture. The area and power of the adder, multiplier, and delay circuits required to realize each filter bank is reported in this work. The modified Mallat structure with the least area power product is suitable for the denoising ECG signal.