Study On ECG Single-fiducial-based Human Identification Algorithm

In the modern times of informatization, digitization and networking, the security of personal information has become more and more important. How to validate and identify a person accurately and conveniently has been paid more attention now, which also has caused great importance to the scholars of the world. The traditional human identification technologies such as IC cards, passwords or keys can no longer adequately satisfy the modern security demands. Biometric identification technology is the most convenient and security identification solution. It refers to the technologies which make use of the inherent and unique physical or behavioral characteristics that come from human beings. These characteristics include fingerprints, face, iris, and voice. Electrocardiogram (ECG) is a biological signal comes from human body in the heart. Recent studies have revealed that ECG can also be used as an efficient identification tool with high security because of its complicated mechanism and unique advantages, that is, ECG cannot be easily forged, stolen, and imitated. Therefore, ECG might have significant applications in the identification field.This paper mainly studies the new human identification algorithm based on ECG signal with single-fiducial detection. Firstly, the ECG signal waveform characteristics and its noise characteristics are analyzed, and a new ECG de-noising method based on translation invariant (TI) wavelet transform and overlapping group shrinkage (OGS) is proposed to eliminate the noise effectively and improve the signal-to-noise ratio. Secondly, the main principle of signal sparse representation is studied and the ECG identification algorithm based on the block sparse Bayesian learning (BSBL) is put forward. Finally, the proposed identification algorithm is verified by the ECG data that come from four public databases and high identification rate is obtained.The main work of this paper is as follows:1. The generation mechanism and morphological characteristics of ECG signal are studied, and the common types of the noise in ECG signal are analyzed, thereby laying a foundation for the noise elimination.2. A new ECG de-noising method based on translation invariant wavelet transform and overlapping group shrinkage is proposed. Firstly, the principle of wavelet transform, wavelet thresholding de-noising and the OGS algorithm are studied systematically. Secondly, discrete wavelet transform is combined with OGS thresholding function to create a new ECG signal de-noising algorithm. Finally, the feasibility and validity of the de-noising algorithm are verified by six kinds of typical analog signals and two kinds of ECG signals that come from different ECG databases.3. A new ECG identification algorithm based on R peak detection and BSBL algorithm is put forward. Firstly, the basic theory of signal sparse representation and several sparse algorithms are systematically studied, and then the principle of signal block sparse representation used for human identification is analyzed. Secondly, the de-noised and normalized test heartbeat is block sparse representated through the BSBL algorithm over an over-complete dictionary consisting of training heartbeats, and the recognition criteria is built according to the block sparse coefficients. Finally, the performance of the proposed ECG identification algorithm are verified by the ECG data come from four public databases. Different parameters can exert certain influences on the recognition results. Thus, the training time, the number of PCA eigenvalues, and the different classification criteria are optimized to obtain the best parameters for the identification system.The research in this paper implements a new kind of biometric technology with simple, robust and high recognition rate, which also provides the theoretical foundation for the research and development of the identification instrument system that based on ECG signal.