基于小波变换和感兴趣区域编码的ECG压缩方法

A Quality Controllable Algorithm for ECG Compression Based on Wavelet Transform and ROI Coding

提出了一种基于小波变换和感兴趣区域编码的ECG压缩方法首先使用正交小波变换对去均值处理后的信号进行多层分解。然后根据对原始信号特征提取的结果,找到感兴趣区域,进而找到与感兴趣区域对应的系数,视这些系数为重要系数而予以保留。对非感兴趣区域系数从小到大排序,根据目标PRDBE(PercentageRoot-mean-squareDifferencewithBaselineEliminated)指标,计算该区域系数阈值并阈值化。通过扫描所有小波系数得到重要系数图。最后对重要系数进行标量量化,对重要系数图进行RLE(RunLengthEncoding)编码,并使用Huffman编码进一步提高压缩比。使用MIT/BIH心律失常数据库测试表明,本方法在最大程度保存诊断信息,获得好的信号质量的同时,也获得了基本满足实际应用需要的压缩比。

This paper presents an ECG compression algorithm based on wavelet transform and region of interest （ROD coding. The algorithm has realized near-lossless coding in ROI and quality controllable lossy coding outside of ROI. After mean removal of the original signal, multi-layer orthogonal discrete wavelet transform is performed. Simultaneously,feature extraction is performed on the original signal to find the position of ROI . The coefficients related to the ROI are important coefficients and kept. Otherwise, the energy loss of the transform domain is calculated according to the goal PRDBE （Percentage Root-mean-square Difference with Baseline Eliminated ）, and then the threshold of the coefficients outside of ROI is determined according to the loss of energy. The important coefficients, which include the coefficients of ROI and the coefficients that are larger than the threshold outside of ROI, are put into a linear quantifier. The map, which records the positions of the important coefficients in the original wavelet coefficients vector, is compressed with a run-length encoder. Huffman coding has been applied to improve the compression ratio. ECG signals taken from the MIT/BIH arrhythmia database are tested, and satisfactory results in terms of clinical information preserving, quality and compress ratio are obtained.