Feeding a Spectrogram or STFT plot of audio to a 2D CNN for classification is one of the approaches to classification. Are there more similar image-like representations?
Dear researcher
I see in this article an important relationship to your question, I hope you read it.
Best regards
Audio-Based Event Detection at Different SNR Settings Using Two-Dimensional Spectrogram Magnitude Representations
https://www.mdpi.com/2079-9292/9/10/1593
You can look for Scalogram
Details are in the link
https://in.mathworks.com/help/signal/ug/scalogram-computation-in-signal-analyzer.html
In our work the system learns Mel Spectrogram representation, in facts, it is a deep neural network that, among other things, learns automatically the mel spectrogram layer focusing on certain frequencies with respect to others.
Code provided. Do not forget to cite our work
https://paperswithcode.com/paper/auco-resnet-an-end-to-end-network-for-covid
Empirical mode decomposition techniques, first invented by Huang of NASA and further refined by a host of others may be worth looking into;
1.) N.E. Huang, Z. Shen, S.R. Long, M.L. Wu, H.H. Shih, Q. Zheng, N.C.
Yen, C.C. Tung and H.H. Liu, “The empirical mode decomposition and
Hilbert spectrum for nonlinear and non-stationary time series analysis,”
Proc. Roy. Soc. London A, Vol. 454, pp. 903–995, 1998.
2.) IEEE SENSORS JOURNAL, VOL. 11, NO. 10, OCTOBER 2011 2565
Empirical Mode Decomposition Technique With Conditional Mutual Information for Denoising Operational Sensor Data
Olufemi A. Omitaomu, Member, IEEE, Vladimir A. Protopopescu, and Auroop R. Ganguly, Member, IEEE
and many more...
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification