Principal Component Analysis (PCA)
Principal component analysis (PCA) is a transformation which allows converting a set of observations possibly correlated into a set of values linearly uncorrelated called principal components. The transformation is defined in such a way that the new orthogonal space exploits the variance of the data. First principal component will have the largest possible variance; the second will have the highest variance under the constraint that it is orthogonal to the first component and so on.
The new basis can be generated based on: Time interval; Events; and percentage of variance.
- Time interval: a piece of the signal will be used to generate the new eigenvectors and eigenvalues to project the original data.
- Events: instead of a piece of signal, several time intervals corresponding with a certain characteristic (for instance, T wave interval) can be used to generate a transformation which exploits the similarity with the selected basis.
- Variance: A percentage of variance is selected and then the algorithm generates the maximum number of new principal components that fits that variance.