Research Of ECG Signal Classification Methods Based On Clustering

ECG(Electrocardiogram) signal is one of the important biological signals in clinical medicine. The accurate automatic analysis and diagnosis of ECG play a key role in cardiovascular disease treatment, and is also an enthusiastic subject by domestic and foreign scholars.First, the research status and some commonly used methods of ECG classification are described in this thesis. Then the research results of cluster analysis are introduced. Later, hierarchy clustering algorithm, the K-means algorithm, fuzzy C-means algorithm and clustering algorithm based on Gaussian mixture model are detailedly studied. The above methods are applied for ECG classification with data in MIT-BIH database. The research finds out that hierarchy clustering algorithm is not suitable for large amounts of data, while the other methods can achieve good classification results. In the research of clustering algorithm based on Gaussian mixture model, the author made a secondary classification to the Gaussian mixture model result by using the K-means algorithm and achieves a much better result.Finally the author has developed an ECG classification experimental platform to verify the correctness of the research results. This research has a realistic significance.