Generalized false discovery rate based adaptive thresholding approach for ECG signal compression

The paper proposes Generalized False Discovery Rate (FDR) based thresholding procedure for estimating wavelet coefficients adaptively to compress ECG signal. To determine the adaptive threshold the connection between thresholding and hypothesis testing is used. In the proposed algorithm False Discovery Threshold (FDT) is achieved by computing the probability of each detail coefficient. The computed probabilities are arranged in ascending order. The adaptive critical significance level is calculated by k-FWER and step up k-FDR procedures of multiple hypotheses testing. The significance levels are compared with computed probabilities to satisfy desired FDR, which provides the FDT. Finally two stage entropy coding is done using zero Run Length Encoding (RLE) followed by Huffman coding. As compared to existing techniques for ECG compression the proposed algorithms achieved very low average value of percentage root mean difference (PRD) at comparable compression ratios (CR). Very less PRD values signifies faithful reconstruction of original signal.