Brief Review — Deep Wavelets for Heart Sound Classification


Sik-Ho Tsang
2 min readJan 1, 2024
Happy New Year 2024 !!! (Free image from Abhinav Sharma)

Deep Wavelets for Heart Sound Classification
GRU, by The University of Tokyo, University of Augsburg, Shenzhen University General Hospital, National Kaohsiung University of Science and Technology, Imperial College London
2019 ISPACS (Sik-Ho Tsang @ Medium)

Heart Sound Classification
2013 … 2023 [2LSTM+3FC, 3CONV+2FC] [NRC-Net] [Log-MelSpectrum+Modified VGGNet] [CNN+BiGRU] [CWT+MFCC+DWT+CNN+MLP]
==== My Other Paper Readings Are Also Over Here ====

  • A novel framework is proposed, which is based on wavelet representations and deep recurrent neural networks (DRNNs) for recognising three heart sounds, i.e., normal, mild, and severe on HSS corpus.


  1. Wavelet+GRU
  2. Results

1. Wavelet+GRU

1.1. Wavelet

‘coif3’ is selected as the wavelet type for extracting wavelet energy features (WEFs) (dimension: 287) from 7 decomposition levels.

1.2. GRU

The DRNN model is the GRU model with three hidden layers (512–256–128).

2. Results

Confusion matrix

As seen above, the model has an excellent performance for recognising the mild class.

  • However, its capacity in classifying normal, and severe needs to be improved. In particular, it is difficult for the current model to distinguish normal and mild, or severe and mild.
  • (This is a very short 2-page paper.)



Sik-Ho Tsang

PhD, Researcher. I share what I learn. :) Linktree: for Twitter, LinkedIn, etc.