The purpose of this article is to diagnose respiratory apnoea in order to help the person avoid further possible risks. In this article, the ECG signal of 70 patients with sleep apnoea in the Physionet database with a sampling rate of 100 Hz is used. Data recording time is 7 to 10 h, the age range is 27 to 60 years, and weighs between 53 to 135 kg. In this article, using electrocardiogram signal processing, the time of occurrence of a respiratory attack on the patient during sleep is predicted. In order to achieve this goal, after generating the HRV signal from the ECG, time and frequency domain properties are extracted from the HRV signal. In the next step, according to statistical analysis, principal component analysis algorithm, and genetic algorithm, the best combination of features is selected in terms of differentiation between two different groups. In order to evaluate the capability of each feature in distinguishing between two attack and non-attack event intervals, the features are compared separately and in combination. The results show that in the HRV signal of people at risk for sleep apnoea, there are features in the vicinity of the attack that distinguish them from times far away from the attack. It was also shown that the feature combination method has a much greater ability to reveal this difference. The results of specificity, sensitivity, and accuracy obtained by combining the features were 99.77%, 97.38%, and 98.25%, respectively, which has a much higher performance than previous studies. Early detection enables the physician and the intensive care unit to take steps to prevent this from happening, which will save the patient's life.
Keywords: Heart rate variability; SVM; genetic algorithm; obstructive sleep apnoea; time and frequency features.