| Human identification or authentication is a serious problem which we must face in our daily life. Compared with the traditional identification methods, such as password, key and other ways, Biometric Identification Technology (BIT) is to use the human body's own physiological or behavioral characteristics for identification. There are many advantages by using BIT, for example, high safety factor, hard to lost and difficult to forget. Therefore it is more and more applied in various areas of the society such as safety protection, case detected, entrance guard, attendance record and medical insurance etc. Currently, a new biometric technology--using electrocardiogram (ECG) features for human identification, has gradually been of interest to many researchers at home and abroad. ECG for human identification is more effective to prevent some shortcomings such as forgery or imitation in fingerprint and voice imitation exist in current BIT system, and it will be a powerful complement for current BIT system.Based on the background and current ECG human identification research, a new method is investigated and some preliminary research and discussion is conducted in this article. Details are as follows:(1) Method of single feature point extracted from ECG.According to the multi-resolution characteristics of wavelet transform and the singularity analysis to clinical ECG signal, decomposition of ECG signal is calculated using atrous algorithm by spline wavelet, and R wave peaks are extracted from the wavelet decomposition coefficient at scale 4. The experiment result shows that it is accurate enough to extract the location of R wave peaks from clinical ECG signals.(2) Method of identification features extracted from ECGBased on the R wave peaks extracted from ECG, QRS and entire cardiac cycle waveforms are extracted as the recognition feature. Because QRS waveform is less affected by change of heart rate, R wave peak point is taken as a reference point by moving forward and backward to extract a fixed-length data as the QRS waveform. But for the entire cardiac cycle waveform, its length is changed with the heart rate and it must include the P-QRS-T waves. The strategy is different. Finally the firm identification features are extracted by average of multiple QRS and cardiac cycle waveforms.(3) Rapid identification.First of all, an identification template database is extracted and built according to the extraction strategy of ECG recognition feature, and then the identifying characteristics of the unknown individual is extracted. According to the result of correlation calculated between QRS waveform of unknown individual and each QRS template, and combining with the threshold value method, the selected candidates range is limited. Then the matching distance between cardiac cycle waveform of unknown individual and each cardiac cycle template in limited range is calculated by Dynamic Time Warping (DTW) algorithm for the optimal match. Finally the smallest match distance is selected as the identification result. Recognition features are selected respectively from single lead and limb leads in the experiment. The experimental results show that DTW algorithm solves the problems when matching two inconsistent length of the two feature vectors caused by change of heart rate. The proposed identification method is effective to combine the advantages of fastness in correlation algorithm and high accuracy in DTW algorithm, and achieve a rapid, accurate identification, while further evidence is shown that using single lead for ECG human identification is feasible.(4) The software's design and realization of ECG human identification.SQL Server 2000 database is selected to create a template platform for ECG human identification, and Visual C++ 6.0 programming language is adopted to realize the system software. Finally the process of identification and the effect of recognition are demonstrated. |
PDF Full Text Request