Brief Review — Recognition of Normal-Abnormal Phonocardiographic Signals Using Deep Convolutional Neural Networks and Mel-frequency Spectral Coefficients
MFSC+CNN, MFSC+CNN J. Physil. Meas.’17
Recognition of Normal-Abnormal Phonocardiographic Signals Using Deep Convolutional Neural Networks and Mel-frequency Spectral Coefficients
MFSC+CNN J. Physil. Meas.’17, by NFQ Technologies LLC, Vilnius; Vilnius Gediminas Technical University
2017 J. Physil. Meas., Over 100 Citations (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 ====
- Deep convolutional neural networks (CNN) and mel-frequency spectral coefficients (MFSC) were used for recognition of normal-abnormal phonocardiographic signals of the human heart.
Outline
- MFCC+CNN
- Results
1. MFSC+CNN
1.1. MFSC
- As shown above, data is partitioned based on class.
- MFSC (MFCC with no DCT) is extracted as features.
- During inference, multiple frames of MFSC are used for prediction and averaging as final answer.
1.2. CNN
- 2 Convolutional layers are used.
- 3 Dense layers are used at later stage.
- Softmax is used at the end.
2. Results
2.1. Ablation Study
The best result (validation accuracy equals to 0.937) was achieved using 256 hidden dense and dropout layers in a second all data reading stage.
2.2. SOTA Comparisons
- The challenge scored entries on the final test set using the depicted techniques are assembled in Table 3.
The current entry for the proposed approach obtained an overall score of 84.15% in the last phase of the challenge, which provided the sixth official score and differs from the best score of 86.02% by just 1.87%.
