Wavelet-based ECG compression using dynamic multi-stage vector quantization

In this paper, an improved wavelet compression algorithm is proposed for electrocardiogram (ECG) signals which is combined with the lifting wavelet transform (WT) and the dynamic multi-stage vector quantization (MS-VQ). The lifting wavelet transformed coefficients in a hierarchical tree order are taken as the components of a vector called a tree vector (TV). Based on the property of wavelet coefficients that gives emphasis to approximation coefficients, the components of a target vector for two stages VQ is differently extracted from different WT subbands. In the first-stage, 32 dimensional TV for crude quantization is extracted in the order of a hierarchical tree from a WT sub-bands except the last sub-band and in the second stage, 64 dimensional code vectors from all WT sub-bands. Each codebook is adaptively updated by the distortion constrained codebook replenishment mechanism (DCCR). The combination of lifting WT and dynamic multi-stage VQ retains feature integrity of the ECG at high compression ratios. Preliminary results indicate that the proposed method excels over previous techniques for high fidelity compression.