The master's thesis presents various functions, tools and applications useful for the analysis and classification of ECG signals using the Matlab program. We discussed procedures for removing noise and trends in data. We looked at the Signal Processing Toolbox package in Matlab and possible ways to extract features from ECG signals. The signals were divided into blocks containing one QRS complex each. To reduce dimensions, we used the PCA method on individual blocks. Using the SVM method, we organized the ECG signals by classifying them into two classes: signals with a normal heartbeat and signals with atrial fibrillation. First, we used the SVM method on the training and test sets, and then we did classification on the training set and used the learned model for prediction on the test set. The obtained results showed that the model has satisfactory accuracy. In order to compare the performance of the model, we finally performed a classification with LSTM neural networks on the reduced data using the PCA method.
|
