ECG Signal Compression Using a Sinusoidal Transformation of Principal Components
Skander Bensegueni, Abdelhak Bennia · International Journal of Software Engineering and Its Applications · 2016
In this paper, we present a compression method of the electrocardiogram (ECG) signals by a discrete cosine transform of principal components in order to reduce the dimensionality of different cardiac records. In first, the principal components analysis is used for high dimensional electrocardiographic data (ECG signal periods) so that we obtain its low dimensional representation by finding orthogonal linear combinations of ECG data with maximal variances. Secondly, we apply a varying discrete cosine transform thresholds of different principal directions of the data to increase the compression rate but preserving a good signal morphology. The results of the simulation on several MIT-BIH database records show that the proposed technique gave good compression ratios without distortions between the original and the reconstructed signal.