Dual-Tree Sparse Decomposition of DWT Filters for ECG Signal Compression and HRV Analysis

Abstract

In this paper, a one-dimensional ECG signal is decomposed as symmetry tree structure at each level using discrete wavelet transforms which outcomes from a larger quantity of insignificant coefficients. They are measured as zero amplitude value and represented as sparse datasets that improve the compression rate, and Huffman coding helps to represent the signal with low bit rate data. These results compressed data codes of large ECG time-series datasets of the signal. Here, different wavelet filters are evaluated for compression based on sparse data from wavelet decomposition. The performance of an algorithm in terms of compression is 43.52% and 42.8% with a 99.9% correlation between original and recovered signals from compressed ECG data for the MIT-BIH arrhythmia and compression dataset, respectively. Further, heart rate variability (HRV) analysis with correlation of R-R intervals in between original and reconstructed ECG signals validates the reconstruction as well as sensitivity of compression technique toward data accuracy.