Research Of Human Identification Algorithm Using ECG Signal Based On Combining Classifiers

At present,there are many methods which were used for identification,such as face recognition,fingerprint identification and so on.With the updated technology,face can be replaced by photographs,fingerprints can be copied.But ECG signal of person is the one and only,can not be copied,ECG signal is mainly used for medical diagnosis and clinical heart disease in recent years,the number of scholars is rising who put forward the research of basing on ECG signal identification,its purpose is to achieve a better and more accurate for human identification,biological recognition technology developed rapidly because of human's own special behavior and physiological characteristics,its reliability and non substitution are very high.In this dissertation,the algorithm of ECG signal identification based on classifier combination is studied,and the QRS waveform is extracted by using feature extraction method without reference point.HOAC-DCT feature extraction,DWT feature extraction,PCA feature extraction and classifier combination algorithm,the proposed method can improve the accuracy of identification.First of all,preprocessing ECG signal is the first work,the original ECG signals are often accompanied by muscle noise frequency interference,frequency interference,baseline drift,because of ECG signal acquisition equipment and other factors,this dissertation uses four order Butterworth band pass filter.Then,the QRS waveform of ECG signals is extracted by using the feature extraction method without the reference point,the HOAC algorithm is used to extract the normalized QRS waveform of ECG signals,and the influence of benchmark points is eliminated.Secondly,the extracted normalized QRS waveform is extracted again by using HOAC,DWT and PCA for identity recognition.Because of the high feature dimension of HOAC,the feature dimension is reduced by DCT algorithm.Each method extract a feature,after finding out three characteristics,using the nearest neighbor classifier classification,finally use the multiplication,the maximum and minimum,median,majority voting rules for the combination of classification results,to find the best combination of ECG signal recognition algorithm,realizing the identification accuracy of ECG signal is higher.Finally,the performance of the proposed algorithm is validated for PTB and MIT-BIH databases,and the accuracy of identification results is obtained by using MATLAB,and the superiority of the proposed algorithm is verified.The experimental results show that in combination rules,the ability of classification combined multiplication and median rule classifier is better than using a single feature extraction,classification ability,the classification error rate is the lowest,to verify the identification of ECG signal classifier combination better classification results based on the method proposed in this dissertation,and is easy to implement and can provide good technical support for system identification based on ECG.