While the general approach is very interesting, the article itself has some scientific shortcomings:
1. There is no information about how have the results produced by the presented algorithms been validated.
2. While some alternative methods are mentioned in the Introduction, a comparison of the method introduced in this article with these previous approaches is missing. The presented work has not been properly discussed in the context of previous work and state-of-the-art.
3. Generally, there are very few references: a) the majority of the Introduction is not supported by references; b) a link to technical details (specification documents) of the transmitter/receiver unit and the microphone would be helpful (if available); c) "the highest frequency found in the actual heart sound" (Chapter 5) is neither specified nor referenced (the actual frequency range remains unclear); d) ... and some other statements could also be supported by references.
Besides these general issues, I propose to reconsider the introduced and discussed use cases:
I doubt that long-term recording is a reasonable use case in adult medicine. I am not an PCG expert, but from the Introduction I understand that PCGs are usually helpful for structural abnormalities. These defects can normally be detected at any time. Long-term recording, on the contrary, is more suitable for detecting sporadic events or long-term variations. However, a quick web search showed that long-term recording seems to be interesting for surveillace of fetal heart sounds, which could be an interesting use case not mentioned yet in the article.
Some minor issues:
- Some ordinate axis units are missing.
- Some greek letters are missing (probably a font conversion problem)
- "frequency response window from 35 kHz to 20 kHz" should probably be "[..] from 35 Hz to 20 kHz".