Noise effect on orthogonal transform compression of ECG signals

Modeling of signals using orthogonal transforms is a very appropriate technique for ECG data compression. An increase of coefficient number (n) increases both signal and noise power reconstruction. A method for selecting the optimum number of coefficients (n*) in noisy records is presented. Adaptive algorithms achieve better coefficient estimation when the input signal is corrupted with uncorrelated noise. In this work we analyze LMS algorithm that improves the estimation of classical inner product. A selection criteria of the mu parameter is obtained in order to outperform inner product performance.