Method of increasing the quality of identification of users of electronic systems based on IH ECG

Blyagoz Sofiet Hazretovna, Shtennikov Dmitry Gennadievich

The article is devoted to improving the accuracy of identification of users of electronic systems based on their electrocardiogram. In this work, an improved algorithm for the functioning of the developed prototype system for user identification based on ECG is used, which differs from the existing ones in the use of three signs of ECG signals. These include: cepstral coefficients, ZCR and entropy. Then they are combined into one input descriptor, which increases the speed of recognition. SVM and implementations of a neural network (multilayer perceptron) using the SciKit Learn and TensorFlow modules were considered as possible options for a decision model for classification. In this case, to improve the quality of the identifier, the result of the classification of all segments of the ECG recording is used.

Keywords:identification, ECG, cepstral coefficients, ZCR, entropy, SVM, neural network.