Advanced Hybrid Information Processing. First International Conference, ADHIP 2017, Harbin, China, July 17–18, 2017, Proceedings

Research Article

Design and Application of Electrocardiograph Diagnosis System Based on Multifractal Theory

Download
195 downloads
  • @INPROCEEDINGS{10.1007/978-3-319-73317-3_50,
        author={Chunkai Zhang and Ao Yin and Haodong Liu and Jingwang Zhang},
        title={Design and Application of Electrocardiograph Diagnosis System Based on Multifractal Theory},
        proceedings={Advanced Hybrid Information Processing. First International Conference, ADHIP 2017, Harbin, China, July 17--18, 2017, Proceedings},
        proceedings_a={ADHIP},
        year={2018},
        month={2},
        keywords={MFDFA Scale-free interval Multifractals Neural networks},
        doi={10.1007/978-3-319-73317-3_50}
    }
    
  • Chunkai Zhang
    Ao Yin
    Haodong Liu
    Jingwang Zhang
    Year: 2018
    Design and Application of Electrocardiograph Diagnosis System Based on Multifractal Theory
    ADHIP
    Springer
    DOI: 10.1007/978-3-319-73317-3_50
Chunkai Zhang1,*, Ao Yin1,*, Haodong Liu1,*, Jingwang Zhang1,*
  • 1: Harbin Institute of Technology Shenzhen Graduate School
*Contact email: ckzhang812@gmail.com, 1065669324@qq.com, Haodong.1994@qq.com, Bluesmile2013@hotmail.com

Abstract

At present there are some ECG automatic diagnosis and identification system, which generally have a common characteristic that their research direction is more inclined to time domain analysis and frequency domain analysis. A large number of researchers have proved that ECG signal has multiple fractal characteristics, while using multi-fractal to analyze the chaotic system is also a trend. In this paper, the main research content is ECG automatic identification: ① Design and implementation of a differential threshold method for ECG signal automatic segmentation algorithm, the algorithm can automatically identify a segment of ECG in the ECG cycle, and ignore those ECG cycles, which are not complete ECG signal. ② Propose an algorithm to describe the data classification by using the multifractal theory to describe the data characteristics. The multi-fractal and semi-spectral characteristics of ECG and generalized Hurst exponent are used to train and test the neural network model. The accuracy of classification is 97%. ③ A complete ECG signal annotation system was built, which can automatically identify a segment of ECG sequence with multiple cycles and annotate each cycle. At the same time it can automatically ignored end-to-end incomplete ECG signal of the ECG sequence, that’s to say, this system has a better fault tolerance.