Study On The Improved Methodology Of ECG Clustering Strategy

Electrocardiogram(short for ECG) can directly show the electrical activityprocess of cardiac excitement，which is an important expression of human vitalsigns．In medical science，diagnosis in allusion to ECG waveform has become animportant link in the detection of heart diseases． Nowadays， there are manydefects and shortages in both theoretical research and in practical application ofautomatic analysis methodologies．As a result，there are still many aspects needimprovement and innovation． In order to improve this situation， the paperresearches on three aspects of the key technology around“pretreatment of ECGwaveform signals”，“detection and localization of ECG waveform signals”， and“automatic clustering of ECG waveform signals”．Pretreatment of ECG waveform signal has been conducted． In order toreduce interference noise such as utility frequency and baseline drift，this paperhas improved traditional wavelet filter for ECG waveform signal and has putforward a methodology that different scales employing corresponding thresholdto process wavelet functions according to actual needs of automatic analysis onECG signals， which has solved the problem that the optimal number ofdecomposition layers in wavelet threshold denoising methodology is difficult todetermine．After the verification experiment by using The MIT-BIH ECG data，it is found that the improved wavelet filter can not only filter out interferencepower frequency and high frequency noise，but also better keep the useful signaldetails in the in the experimental verification．An improved algorithm for implementation of detecting QRS wave based onwavelet transformation is proposed in this paper，which is used for improvementof traditional wavelet transform and thus overcome defects such as complexity incomputation，slowness in detection speed，and easiness to misjudge the pseudodifferential signals．According to the difficulty to detection and localization of P wave and T wave in ECG waveform signals，the idea of self-adaptive cancellationbetween each other has been applied to QRS-T waveform self-adaptivecancellation between each other and to increase the proportion of P wave in therest of the waveform．The detection methodology that could well eliminate theeffects of high frequency noise and baseline drift，and could not be influenced bywave shape of QRS and the occurrence of arrhythmia performs excellent in thedetection of P wave．In order to improve automatic clustering of the ECG waveform signal，thispaper proposes a new clustering algorithm which is an improved K-meansclustering algorithm based on mean square error weighted genetic simulatedannealing to cluster ECG signals． The simulation experiments have beenconducted by employing typical data from MIT-BIH standard ECG database．That the Accuracy of K-means clustering algorithm improved in this paper isapparently superior to the traditional one proved this algorithm is effective tocluster ECG signals．...