Title: Adaptive thresholding of wavelet coefficients using generalised false discovery rate to compress ECG signal

Authors: Supriya O. Rajankar; Sanjay N. Talbar

Addresses: Sinhgad College of Engineering, Pune 411041, India ' Shri Guru Gobind Singhji Institute of Engineering and Technology, Nanded 431606, India

Abstract: In signal compression the selection of an appropriate threshold is the challenging task. The paper proposes an algorithm to determine the signal adaptive threshold based on estimating wavelet coefficients by generalised False Discovery Rate (FDR) to compress ECG signal. The hypothesis testing and thresholding are closely related. So, multiple hypotheses testing is used to determine an adaptive threshold called as False Discovery Threshold (FDT). The p-value of each wavelet detail coefficient is computed and arranged in an ascending manner. The dynamic critical significance levels are calculated using k-FWER and k-FDR. These significance levels are compared with the corresponding p-value to satisfy desired FDR, which provides the FDT. The run length encoding followed by Huffman coding provides compression. The paper also proposes a new performance evaluation parameter: mean Structural Similarity Index (mSSIM) to check the similarity between original and reconstructed ECG signals. Generalised FDR-based thresholding provides very less PRD value compared to standard codecs in the literature and structural similarity very close to one, which signifies better reconstruction of the signal.

Keywords: generalised false discovery rate; step-up procedure; BH procedure; k-FDR; k-FWER.

DOI: 10.1504/IJBET.2019.097303

International Journal of Biomedical Engineering and Technology, 2019 Vol.29 No.2, pp.155 - 173

Received: 08 Apr 2016
Accepted: 25 Nov 2016
Published online: 14 Jan 2019 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article