In this paper, we present a new method to categorize the heart sound signals into Normal, Murmur, Extra Heart Sound and Artifact with Convolutional Neural Network (CNN) model. The experimental dataset are from the PASCAL heart sound classification workshop, 2011. Our method mainly contains three steps as follows. Firstly, we propose cross- cutting and framing to enlarge the data. The data used for training and test is three times than before. Secondly, we manage the heart sound signals via Butterworth filter and Fast Fourier Transform simply. Finally the captured frequency features are input into the CNN model for classification. Compared with the traditional heart sounds classification approaches-selecting features and designing classifier, our method overcomes the uncertain in selecting features and improves the self-learning ability. The experimental results show that deep learning network model have higher classification effectiveness for the data with much noise. The global identification rate reaches 98%, more effective than previous related studies.
[1]
Nitish Srivastava,et al.
Dropout: a simple way to prevent neural networks from overfitting
,
2014,
J. Mach. Learn. Res..
[2]
Shi-Wen Deng,et al.
Towards heart sound classification without segmentation via autocorrelation feature and diffusion maps
,
2016,
Future Gener. Comput. Syst..
[3]
Prospero C. Naval,et al.
Classification of heart sounds using discrete and continuous wavelet transform and random forests
,
2015,
2015 3rd IAPR Asian Conference on Pattern Recognition (ACPR).
[4]
Wenjie Zhang,et al.
Heart sound classification based on scaled spectrogram and partial least squares regression
,
2017,
Biomed. Signal Process. Control..
[5]
Mohamed Esmail Karar,et al.
Automated Diagnosis of Heart Sounds Using Rule-Based Classification Tree
,
2017,
Journal of Medical Systems.