An investigation of adaptive beat estimation for improved data compression

The authors have investigated the use of adaptive beat estimation for ECG data compression. On selected MIT/BIH database records, they estimated QRS complexes from previous QRS complexes with simple averaging, LMS adaptive estimation, and averaging based on the mean axis direction of the QRS complex. The authors subtracted beat estimates from actual beats to produce residual beats. They calculated entropy of the first differenced residual data. Differenced residual signal entropy was reduced by adaptive estimation in thirteen of fourteen records, with a maximum reduction of 0.53%. Averaging based on mean QRS axis reduced the differenced residual signal entropy in 11 of 13 records, with a maximum entropy reduction of 4.1%.