I am working on a problem statement where I need to extract the relevant part/section from audio files. for instance my data has audio files of engines with background noise - birds, people conversation, etc.
I read couple of papers which have mainly used PCA for denoising the noise. Couple of them have used SNR too. But I am not very convinced as how this can actually work considering that PCA take n-components value so how we can be sure that the components that it is picking up will actually not contain noise?.
The articles/links that I explored for this are:
https://arxiv.org/pdf/1804.07177.pdf (applies SNR),
https://refractedrendition.wordpress.com/ + http://willdrevo.com/fingerprinting-and-audio-recognition-with-python/ + https://www.acrcloud.com/docs/acrcloud/introduction/audio-fingerprinting/ + http://www.cp.jku.at/research/papers/Sonnleitner_Dissertation.pdf + http://www.music.mcgill.ca/~ich/classes/mumt621_09/fingerprinting/cano05review.pdf +n http://www.ismir2002.ismir.net/proceedings/02-FP04-2.pdf + http://www.ismir2002.ismir.net/proceedings/02-FP04-2.pdf (audio-fingerprinting),
Any lead as to how this relevant part extraction can be done would be very helpful to me.
Looking forward to getting good leads in solving this problem.