Heart murmur classification using MFCCs


I’m working on a project to classify and predict heart murmurs. The project is at the initial phase. Basically, the preprocessing is critical when considering the denoising and segmentation. Recorded heart sounds (PCGs) are interpreting some of the pathological insights of cardiac vascular diseases. According to my initial findings, spectrogram based CNNs are popular in the domain. What would be more appropriate architectures to build a classification model? Highly appreciate your comments.

Thank you!

Hi AsiriSirithunga,

I just came across your post. I do not know whether this is still useful to you, but it may be interesting for your case to look into the use of transformers. You can find examples in this and this paper.

1 Like