Research On Clustering In Typical Character Of ECG By Fussy C-Means

As sciences and technologies are under vigorous development, data size are sharply growing, both the timing and the complexity of data are beyond the capability of the technology of data processing and knowledge discovery. Nowadays we have this is a brilliant characteristic of 21th century and which posed a great challenge to the development of our civilization. The so called "data explosion" or "information explosion" is emerged because the capability of data processing is dwarfed by those of data producing and data collecting. Thus there are requirements to abandon redundant data and expect to relieve from processing and analyzing data on low level, then acquire knowledge on a relative high level.We want to make it clear that even data are extremely plentiful, those useful are in little proportion especially under certain conditions. It is a subject full of significant meaning to draw useful rare data from data with redundancy and then to obtain novel knowledge by them.Clustering is an effective method to reduce the complexity of data, and a powerful measure to transfer rare data into knowledge, it is applied successfully in many domain such as biological sciences, medicines, financs, telecom, commerce and scientific researches etc. Rare data refer to those data have rare proportion in whole data while they contain vast useful information, hence is valuable for research. It is a important principle to judge whether a algorithm is effective or not by it's capability to turn rare data into knowledge.Fuzzy clustering algorithm tries to simulate the way human recognizing patterns and acquiring knowleges. Fuzzy clustering algorithm is a combination of fuzzy mathematics and clustering, it allows each sample not has to attach to a certain class during clustering process and the relation between sample and class could be vague. While most traditional clustering algorithm claim every sample should absolutely attach to a certain class, which make the fussy clustering algorithm flexible and immune to many disturbances, thus the fussy clustering is more adaptive to pattern recognize of rare data.The fuzzy-clustering algorithm based on our investigation takes the famous MIT/BIH database as the means to test clustering precision and the performance of detecting rare data. It is made up of three modules—the QRS location searching module, the character detecting module and ECG information clustering module as well.In the QRS location searching module, we didn't adopt wavlet detecting algorithm, however we combine difference detect technology with statistic method which eventually provide quick and accurate QRS location searching performance with good accuracy after compare with MIT comment. In the character detecting module, we picked out 26 typical characters from each QRS wave which form a vector used by clustering module to evaluate distance between each sample. And in the clustering module, we adopted the structure of add clustering combined with FCM(fuzzy c-means), then we find with all those above efforts we can achieve high precision and good performance of rare pathological data classification detecting.The thesis was consist of five chapters as follows:Chapter one introduces the background, arrange of this project and presentation of studying problem as well.The second chapter was fuzzy clustering analysis.The third chapter design fuzzy clustering algorithm.The fourth chapter presents an enssicial characters of electrocardiogram signal.The fifth chapter talks about applications of fuzzy clustering algorithm, and gives an analysis on the result.Finally, we take a look on the deficiency of the algorithm and give some viewpoint about imrpving in futre.