Compression of the ambulatory ECG by average beat subtraction and residual differencing

A strategy is evaluated for compression of ambulatory electrocardiograms (ECGs) that uses average beat subtraction and Huffman coding of the differenced residual signal. A sample rate of 100 sps and a quantization level of 35 mu V are selected to minimize the mean-square-error distortion while maintaining a data rate that allows 24 h of two-channel ECG data to be stored in less than 4 MB of memory. With this method, sample rate, and quantization level, the ambulatory ECG is compressed and stored in real time with an average data rate of 174 b/s per channel. It is demonstrated that, for the low sample rate and coarse quantization required for ambulatory recording, without sufficient temporal resolution in beat location, beat subtraction does not significantly improve compression, and may even worsen compression performance. It is also shown that with average beat subtraction, compression is improved if multiple beat averages maintained. Improvement is most significant for ECG signals that exhibit frequent ectopic beats.