ECG signal compressed sensing using the wavelet tree model

Abstract

Compressed Sensing (CS) is a novel approach of compressing, which can reconstruct a sparse signal much below Nyquist rate of sampling. Though ECG signals can be well approximated by some wavelet basis, the noise still influences the ECG wavelet decomposition and also reduces the effectiveness of the signal reconstruction. In this note, we present a compressed sensing method to reconstruct ECG signals in MITBIH [1] arrhythmia based on different wavelet families. Our approach is composed of two steps. The first step is to use Condensing Sort and Select Algorithm (CSSA) to dampen the impact of the noise for ECG signals and get sparse signals to estimate and replace raw ECG signals, and then, the second step is to use CS method to compress and transfer those filtered signals. The result is evaluated by some indices like Percentage Root Mean Square Difference (PRD), Mean Square Error (MSE), Peak Signal To Noise Ratio (PSNR), and Correlation Coefficient (CoC). These reconstructed results are comprehensively compared by 4:1 compression ratio. These results indicate that Symlets and Daubechies wavelet families have better performance for all parameters compared to other wavelet families and most of existing results.