Open AccessResearch on Arrhythmia Classification Method Using Optimized Probabilistic Neural Network
Author(s) -
Liu Jun-kong,
Lili Li,
Fang Yi-pin,
Su Jinyu
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1939/1/012104
Subject(s) - pattern recognition (psychology) , artificial intelligence , computer science , artificial neural network , probabilistic neural network , probabilistic logic , similarity (geometry) , feature (linguistics) , dimension (graph theory) , feature extraction , signal (programming language) , standard deviation , cardiac arrhythmia , data mining , mathematics , statistics , time delay neural network , medicine , linguistics , philosophy , pure mathematics , programming language , cardiology , image (mathematics) , atrial fibrillation
Aiming at the classification difficulty of complex and diverse ECG signals, this paper proposes a feature extraction method based on standard deviation. This method solves the problems of multi-dimensional, diverse, high similarity and the difficulty in extracting main features effectively of ECG(electrocardiogram, ECG) signal features. In addition, this method overcomes the difficulty of low classification accuracy because of large differences in ECG signals of the same type among different patients. This paper adopts optimized probabilistic neural network methods to achieve automatic classification of arrhythmia. First, the standard deviation of the dimensions of each sampling point of the ECG signal would be calculated and sorted by size. Second, the first m dimensions as the feature dimensions of arrhythmia would be extracted. After that, the probabilistic neural network would be used to train and classify feature dimension data. Finally Bayesian Optimization(BO) method would be used to optimize the parameters globally. In the experiment of the MIT-BIH arrhythmia database, the arrhythmia data was divided into 5 categories and verified, experimental results show that the correct rate of classification of the arrhythmia data of patients reached 99.67%, which proved the effectiveness of the method in this paper.