Study of Dynamic Electrocardiogram Waveform Features Clustering by Wavelet Transform
Pengtao Wang · Journal of Optoelectronics·laser · 2007
The paper mainly studied the electrocardiogram wave cluster of dynamic electrocardiogram,the clustering result will find the basic wave.The study used quardratic spline wavelet to detect R wave in cardiogram,and confimed the wave of an whole heart period.At same time the maximum and minimum points,related slopes were obtained as the features extractioin vectors.These vectors were clustered by self-organization map(SOM) neural netowrk.Through the experiment,the R wave detection rate was up to 99.5%,which is better than those on artificial neural network and linear filter.The basic recognition rate of 24 hours dynamic electrocardiogram,which includes 100,000 cadiogram waves,reaches 91.4%.The results reduce the cardiogram data 5～10%,which is helpful to doctor's diagnosis.