Wavelet-Based Quality-Constrained ECG Data Compression System Without Decoding Process

Abstract

Retaining the reconstruction quality of electrocardiogram (ECG) data compression is crucial for diagnosis demands. However, traditional wavelet-based approaches suffer from complicated distortion control. The most fundamental way to overcome this difficulty is to avoid the use of the time-domain distortion measure that requires the complete decoding process. In this article, based on the reversible roundoff nonrecursive discrete periodized wavelet transform reversible round-off non-recursive discrete periodized wavelet transform (RRO-NRDPWT), a new transform-domain index is presented. This index accumulates the roundoff error of the RRO-NRDPWT and the quantization error as the distortion measure. The index referred to as the simplified percentage root-mean-square difference (SPRD) can completely replace the PRD for the case of ${\mathbf{PRD}} \geq 2{\mathrm{\% }}$PRD≥2%, where $\boldsymbol{SPRD}/\boldsymbol{PRD} \approx 1$SPRD/PRD≈1. By using the ECG signals in the MIT-BIH arrhythmia database, the simulation results show that for target PRD (${\mathbf{PR}}{{\mathbf{D}}_{\mathbf T}}$PRDT) equals to $3{\mathrm{\% }} \pm 0.05$3%±0.05, the SPRD can save 99.34% quantization process time compared with the traditional quality control scheme.