Canonical correlation analysis for dimensionality reduction of sleep apnea features based on ECG single lead

This paper presents the canonical correlation analysis (CCA) method for dimensional reduction of sleep apnea features extracted from the electrocardiogram single lead. The feature extraction belong to the linear and nonlinear techniques of variance of heart rhythm or heart rate variability and respiratory waveform from electrocardiography signal. These are benefit to evaluate the sleep apnea in noninvasive way. We introduce a feature reduction using the CCA method to determine relationship across the pair data sets and also compare with the common dimensionality reduction methods. As our review, this first demonstrate offers the CCA method to transformed apnea features. In experiment, 25 recording from physionet database are used and all features are extracted minute-by-minute based on apnea annotation. As expected, in processing for CCA, the classification results are better than the classical methods, and use the number of reduced features less than the classical ones.