Research On Electrocardiogram Identification With Multi-dimensional Features

In recent years, the analysis and process technology of multi-dimensional data is growing concerned with the rapid development of science and technology, biomedical, Internet, security cerfification and financial fields and so on. However, the traditional one-dimensional features algorithms which for analyszing and processing multi-dimensional data is difficult to apply to multi-dimensional data directly. Electrocardiogram, as an important human identification biometric security feature, which make it more complex to process and analysis with its multi-dimensional characteristics. Most of the ECG biometric researchs presented using only one single-lead information. Now this paper proposed a multi-dimensional feature identification algorithm with electrocardiogram. We study on its feasibility and safety based on multiple leads signals of ECG.Fusing multi-dimensional characteristic variables can provide a full and comprehensive views to do identification research. This paper proposed three basic multi-dimensional feature recognition algorithms according to the multi-dimensional features study of the algorithm in depth and on the basis of existing studies and experiments. First, sparse matrix correlation coefficient algorithm based on the MIT-BIH databases, which mapping two-leads signals into two-dimensional coordinates and then extract sparse features to calculate correlation coefficient to identify human. Second, the algorithm based on self-collected ECG and blood oxygen signals data. After mapping signals data into two-dimensional, quantifying sparse matrix characteristics and then to identify. The third algorithm is based on PTB multi-dimensional databases, and using sparse DTW to improve recognition result and space efficiency. And the result of our experiments demonstrate their recognition can reach 96.9%, 93.14% and 98.67%. While effect of the second algorithm of quantization sparse feature-based wasn’t as good as other algorithms, it is our own acquisition, and it is real and reliable, and a strong practical. This results in the real recognition is acceptable.Those three algorithms of this subject to verify the results of our human identification from different databases. And each database with its own advantages and disadvantages. But they are all prove the correctness and availability of our multi-dimensional feature with ECG from different perspective. They optimize algorithms and proposed new framework step by step further. And the recognition result can be more than 10% compared with other alogrithm s ultimately. The multi-dimensional feature recognition algorithm framework have relatively high reference value for subsequent studies.