Multi-modal Feature Based for Phonocardiogram Signal Classification Using Autoencoder

  • Muhamad FathurahmanEmail author
  • Ummi Azizah Rachmawati
  • Sri Chusri Haryanti
Conference paper
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 978)


Phonocardiogram classification plays an important rule in the diagnosis of heart disease. It can be used in selecting a proper treatment to the patients. However, automated PCG classification has many issues. One of the important issues is the feature extraction process. It is difficult to extract relevant features from PCG signals due to some noises that corrupt the original signal. The noises are included murmur, intestine and breathing sounds. To overcome this problem, several works have been proposed such as performing segmentation on PCG signals before the feature extraction process, using many types of signal features including wavelet, mfcc, spectral, time-frequency and statistical features. These types of features experimentally affect the classification accuracy of PCG signals. This study proposes a feature fusion based using an autoencoder model in order to obtained new repfresentation features. The result shows that this approach provides a competitive result of PCG classification compare to those of the baseline methods.


PCG classification Autoencoder Feature fussion Heart sounds 



This research is supported by Internal Research Grant of Universitas YARSI Number: 0007.2/FTI/ST-PN.00/I/2019.


  1. 1.
    Saini M (2016) Proposed algorithm for implementation of Shannon energy envelope for heart sound analysis. 7109(3):15–19Google Scholar
  2. 2.
    Deng Y, Bentley P, A robust heart sound segmentation and classification algorithm using wavelet decomposition and spectrogram. Peterjbentley.Com
  3. 3.
    Springer DB, Tarassenko L, Clifford GD (2016) Logistic regression-HSMM-based heart sound segmentation. IEEE Trans Biomed Eng 63(4):822–832Google Scholar
  4. 4.
    Imani M, Ghassemian H (2016) Curve fitting, filter bank and wavelet feature fusion for classification of PCG signals. In: 2016 24th Iran. Conf. Electr. Eng. ICEE 2016, pp 203–208Google Scholar
  5. 5.
    Tschannen M, Kramer T, Marti G, Heinzmann M, Wiatowski T (2016) Heart sound classification using deep structured features. Comput Cardiol Conf (CinC), 2016Google Scholar
  6. 6.
    Leal A et al (2018) Noise detection in phonocardiograms by exploring similarities in spectral features. Biomed Signal Process Control 44:154–167CrossRefGoogle Scholar
  7. 7.
    Cheng X, Sun K, Zhang X, She C (2016) Feature extraction and recognition methods based on phonocardiogram, pp 87–92Google Scholar
  8. 8.
    Randhawa SK, Singh M (2015) Classification of heart sound signals using multi-modal features. Procedia Comput Sci 58:165–171CrossRefGoogle Scholar
  9. 9.
    Zhang W, Han J, Deng S (2017) Heart sound classification based on scaled spectrogram and partial least squares regression. Biomed Signal Process Control 32:20–28CrossRefGoogle Scholar
  10. 10.
    Vachhani B, Bhat C, Das B, Kopparapu SK (2017) Deep autoencoder based speech features for improved dysarthric speech recognition. In: Proceedings annual conference international speech communication association INTERSPEECH, pp 1854–1858Google Scholar
  11. 11.
    Deng J, Zhang Z, Eyben F, Schuller B (2014) Autoencoder-based unsupervised domain adaptation for speech emotion recognition. IEEE Signal Process Lett 21(9):1068–1072CrossRefGoogle Scholar
  12. 12.
    Oldˇrich Plchot PM, Burget L, Aronowitz H (2016) Audio enhancing with DNN autoencoder for speaker recognition, pp 5090–5094Google Scholar
  13. 13.
    Peter Bentley RG, Nordehn G, Coimbra M, Mannor S (2012) Classifying heart sounds challenge. [Online]. Available: Accessed 09 Apr 2019
  14. 14.
    Bentley P, Nordehn G, Coimbra M, Mannor S (2011) The PASCAL classifying heart sound challenge 2011 results. [Online]. Available: Accessed 16 Jun 2019
  15. 15.
    Guorong W, Dinggang S, Mert SR (2016) Machine learning and medical imaging. ElsevierGoogle Scholar
  16. 16.
    Brosch T, Yoo Y, Tang LYW, Tam R (2016) Deep learning of brain images and its application to multiple sclerosis, no. 1. Elsevier IncGoogle Scholar
  17. 17.
    Atibi M, Atouf I, Boussaa M, Bennis A (2016) Comparison between the MFCC and DWT applied to the roadway classification. In: Proceedings—CSIT 2016 2016 7th international conference on computer science information technology, pp 0–4Google Scholar
  18. 18.
    Extract mfcc, log energy, delta, and delta-delta of audio signal—MATLAB mfcc. [Online]. Available: Accessed 16 Jun 2019
  19. 19.
    Kumar D, Carvalho P, Antunes M, Paiva RP, Henriques J (2011) An adaptive approach to abnormal heart sound segmentation, pp 661–664Google Scholar
  20. 20.
    Maximal overlap discrete wavelet transform—MATLAB modwt. [Online]. Available: Accessed 19 Jun 2019
  21. 21.
    Multiscale variance of maximal overlap discrete wavelet transform—MATLAB modwtvar. [Online]. Available: Accessed 19 Jun 2019
  22. 22.
    Matlab Documentation, Support Vector Machine Classification—MATLAB & Simulink (2019). [Online]. Available: Accessed 09 Apr 2019
  23. 23.
    Marques N, Almeida R, Rocha AP, Coimbra M (2013) Exploring the stationary wavelet transform detail coefficients for detection and identification of the S1 and S2 heart sounds. Comput Cardiol Conf (CinC), 891–894Google Scholar

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  • Muhamad Fathurahman
    • 1
    Email author
  • Ummi Azizah Rachmawati
    • 1
  • Sri Chusri Haryanti
    • 1
  1. 1.Fakultas Teknologi Informasi, Teknik InformatikaUniversitas YARSIJakartaIndonesia

Personalised recommendations