Data compression applied to dynamic electrocardiography

We have tested some techniques of ECG compression on the BIH/MIT Arrhythmias database. We have applied the following methods: (1) method of differences; (2) compression by prediction; (3) sample skipping methods; with the following kinds of errors: amplitude error, amplitude and delay error, area error. In the first and second method the average code length found after application of the Huffmann encoding has been found to be about 4 bits sample-1. For the third method the final compression ratio varies according to the allowed error; with a limited error threshold, however, we have reached an average compression a little lower than 1:5.