ECG compression based on empirical mode decomposition and tunable-Q wavelet transform with validation using heartbeat classification

In telemedicine-based healthcare system, such as cardiac health monitoring system, large amount of data needs to be stored and transferred. This requires stupendous bandwidth and affects the channel efficiency. The main objective is to develop an efficient compression technique for solving such problems in healthcare systems. In this work, the coalition of empirical mode decomposition (EMD) and tunable quality wavelet transform (TQWT) scheme has been proposed for ECG signal compression with a suitable decomposition level. Thus, the maximum energy is packed for fewer coefficients which have a significant contribution to the original signal. The dynamic thresholding and dead-zone quantization are evaluated, to discard the wavelet coefficients with a small value near zero. Subsequently, a run-length encoding (RLE) lossless compression scheme is employed to encode the wavelet coefficients. The presented technique was evaluated on the Massachusetts Institute of Technology-Beth Israel Hospital (MITDB) arrhythmias dataset which contain regular and irregular heart rhythm. The compression ratio (CR%), percent root-mean-square error (PRD%), normalized PRD (NPRD%), quality score (QS), and signal-to-noise ratio (SNR) of 33.11, 4.35, 8.21, 7.59, and 51.09 have been achieved, after implementing on 48 ECG records with 30-min duration. The presented method was also implemented for normal and abnormal heartbeat classification for validation. The random forest algorithm (RFA) is employed for the classification of cardiac rhythm. The results show minimal distortion with an improved reconstruction of a signal after the compression and show a better performance than the state-of-art technique.