Sometimes noise is signal, a recording of the wind or waves crashing on rocks, but I wonder if there wouldn't be some merit to a tool that measured the content of a file with no harmonic relationship to the predominant signal?
A quick and dirty estimate might come from looking at some average level of the quietest parts that are not artificially set to zero or some dither value. Between notes after decay ends, there might not be a lot of such spots, but they might have some common level that could be tracked.
I'm not sure what use knowing the noise floor might be, but I think it might be interesting.
Are you familiar with lossywav? This is essentially what it does.
Are you familiar with lossywav? This is essentially what it does.
Thanks, not til now, just looked it up. I haven't read enough yet to see if it generates any kind of numerical assessment, or if that would require adding on something. I'll be studying it once I am awake again, and post again if I have questions.
If you want the raw data, you can simply process a file and compare the difference between the original and the processed version. Thats essentially a very good estimate of the noise power per sample.
Sometimes noise is signal, a recording of the wind or waves crashing on rocks, but I wonder if there wouldn't be some merit to a tool that measured the content of a file with no harmonic relationship to the predominant signal?
You've got a serious inherent problem. The relationship between fundamentals and overtones produced by many musical instruments aren't exact integer relationships.
I usually judge the noise in a file by the amplitude of the file when there is no music playing. However many realk world noise sources such as HVAC equipment that include large amplitude signals with harmonic relationships.
So you've got a conundrum with using the presence or absence of harmonic relationships as a guide. Their presence is not reliably indicative of the signal being music, and their absense is not reliably indicative of the signal being noise.