Lossy-to-Lossless Compression for Multichannel ECG Using Two Dimensional Transform and Successive Approximation Coding

• Main Authors: Tsang-Chih Wu, 吳倉志
• Other Authors: Shaou-Gang Miaou
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/08828845996453448400

碩士 === 中原大學 === 電子工程研究所 === 93 === The traditional Electrocardiogram (ECG) has twelve leads, including six frontal plane leads and six horizontal plane leads, and the six frontal plane leads (from V1 to V6) correlate closely with each other. In the past, lossy and lossless compression methods are often proposed for single channel ECG signals, and compression methods that are directly applied to multichannel ECG signals are rarely seen. For multichannel ECG signals, encoding ECG signals on a channel by channel basis is not efficient because the correlation across channels is not exploited. We propose a new ECG compression approach using a two-dimensional (2D) integer wavelet transform (IWT) and the set partitioning in hierarchical trees (SPIHT) along with successive approximation coding (SAC). The 2D-transform exploits both the in-channel and the inter-channel correlations. The transform results in wavelet coefficient blocks. Whether a block will be sent is determined by a block significance classifier. The IWT guarantees a true lossless compression, and the SPIHT allows efficient coding and progressive transition from lossy to lossless compression. The ECG data of 8 leads in the CSE Database are tested. The experimental results show that the proposed approach performs better than its single-channel version in both lossy and lossless cases. For lossy compression, when compression ratio is 12.34, the average PRD is 7.56% using the single channel and 3.21% using the proposed approach. The performance of the proposed approach is significantly better than that of the single channel version. For lossless compression, the average compressed data rate per channel is reduced from 6000 bits/s to 3007.8 bits/s using the single channel approach and to 2876 bits/s using the proposed approach.