In this paper, we describe a multi-lead electrocardiogram (MECG) compression technique, which preserves pathological information in different affected leads while achieving high overall compression. For non-affected leads, the principal component decomposed expansion coefficients were optimally quantized using a feed-forward neural network. For affected leads, the wavelet decomposed coefficients were quantized using a fixed level. The proposed technique was evaluated with 130 ECG records with three major classes of myocardial infarction under Physionet. An average overall compression ratio of 21.25, with low values of percentage root mean squared difference of 2.45 for the affected lead group, was obtained.

在本文中，我们描述了一种多导联心电图 ( M ECG) 压缩技术，该技术保留不同受影响导联的病理信息，同时实现高整体压缩。对于未受影响的导联，使用前馈神经网络对主成分分解膨胀系数进行优化量化。对于受影响的导联，小波分解系数使用固定水平进行量化。所提出的技术在 Physionet 下用 130 条心电图记录评估了三种主要类型的心肌梗塞。获得了 21.25 的平均整体压缩比，受影响的铅组的百分比均方根差值较低，为 2.45。