Research Of The Automatic Selecting Strategy Of Typical ECG Waveform For Identification

Identification is particularly important in the field of security.The traditional identification technology is widely used in the field of security,forget.Electrocardiogram(ECG)signals is difficult to be stolen and copied.Compared with other traditional biological characteristics,ECG signals is more safety.So,it has been used in the field of human identity recognition.There are two main methods of ECG identity recognition: one is based on reference point and the other is based on waveform.No matter which method is used for identification,the establishment of template library is needed.The quality of data in template library directly affects the accuracy of identification.ECG is a kind of weak electric signal.It is easy to be disturbed by acquisition equipment and human activities in the process of acquisition.Therefore,ECG waveform selecting has some practical significance.Aiming at the time-consuming problem of manual selection of waveforms used in the establishment of identification templates and the lack of standardization and typicality,a typical ECG waveform automatic screening strategy for identification is proposed in this paper.The randomly selected single cycle waveform is not complete(such as the absence of P waves,T waves are absent,R wave loss).In order to solve the problems of deformation and affect the identification accuracy of ECG identity,we proposed a selecting method of standard ECG waveform based on Gauss mixture model method.The standard ECG signal gathered by ECG signal generator is cut into a single cycle.And 46 features of amplitude,slope,kurtosis,arc length and area of each cardiac cycle were extracted,which are used to establish the Gauss mixture model of single cardiac cycle standard ECG waveform.By calculating Mahalanobis distance to judge whether a ECG waveform accords with Gauss distribution model of standard ECG waveform,the ECG waveform selecting with obvious PQRST characteristics can be realized.The method was tested in 52 people from six different data sets,the average correct rate of selecting was 85.91%,results show that the standard complete ECG waveform could be selected effectively.In order to make the selected ECG waveform more typical,this paper construct a typical Gaussian model of P-wave,QRS complex wave and T-wave respectively.The selecting strategy was used to select 52 individuals' original data,and the identification experiment was carried out.The traditional correlation coefficient method,Euclidean distance,Angle cosine method,Manhattan distance and cosine theorem are used for identification on the same experimental data set.The experimental results show that: the average correct rate of selecting was 94.6% and reject rate of incomplete waveform is 98.52;by using proposed method,the accuracy of identification was increased by 11.79%,10.06%,15.89%,11.44% and 16.3%,respectively.