Reduction of Walsh-transformed electrocardiograms by double logarithmic coding

This work presents an electrocardiogram ECG data reduction method based on Walsh spectrum double logarithmic quantization. The technique is theoretically justified for a simulated ECG and its practical efficiency confirmed using MIT/BIH arrhythmia database signals. By classifying a "good" compression as one with MSE/spl les/0.005 for 1:1 spectral reduction, a normal/abnormal ECG mix returned an 87% success rate for waveforms stored with 8- to 11-bit resolution.