Abstract

The removal of random noises in electrocardiogram (ECG) using adaptive noise canceller (called single-input adaptive noise canceller) without reference input is presented in this paper. Common approaches for noise cancellation require reference input that must be well-correlated with the noise part of the primary input. However, the reference input may be limited in availability and hence, results in degradation of performance. ECG signals can be treated as quasi-periodic signals relative to their additive random noises. This paves the way for the possibility of using single-input adaptive noise canceller for the removal of random noises in ECG under limited availability of reference input.  Computer simulation results verified that commonly used adaptive noise canceller cannot perform well for a poorly correlated reference input. Also, the results indicated that the single-input adaptive noise canceller with delays in the primary input performs almost the same as the commonly used adaptive noise canceller under a well-correlated reference input.

Key words: Adaptive noise canceller, electrocardiogram (ECG), mean-square error, reference input, single-input adaptive noise canceller.