pyAudioAnalysis VS dejavu

I actually came up with an el cheapo version of what I want to accomplish that isn't perfect but without any research can implement it and it may actually prove useful to language learners. PM me if you're interested in hearing it and critiquing it. I can share here that I'm using this guy's multiple repos though: https://github.com/tyiannak/pyAudioAnalysis

Update: From one researcher to another. I was referred to a Python Audio AI project . Once I determine exactly which module to use I should be smooth sailing. I'll send more updates soon.

Hey folks, I'm looking into clustering audio files with features extracted by pyAudioAnalysis. However, every feature (I'm interested in MFCC, spectral centroid and spread, and BPM) is extracted for each frame of the song (by default 0.05s, excluding BPM that relates to the whole) so tracks with different lengths produce arrays with different shapes.

To learn more about pyAudioAnalysis here you go.

Humans are great at classifying noises. We can hear a chirp and surmise that it belongs to a bird, we can hear an abstract noise and classify it as as speech with a particular meaning and definition. This relationship between humans and audio classification forms the basis of speech and human communication as a whole. Translating this incredible ability to computers on the other hand can be a difficult challenge to say the least. Whilst we can naturally decompose signals, how do we teach computers to do this, and how do we show what parts of the signal matter and what parts of the signal are irrelevant or noisy? This is where PyAudio Analysis comes in. PyAudio Analysis is an open source Python project by Theodoros Giannakopoulos, a Principle researcher of multimodal machine learning at the Multimedia Analysis Group of the Computational Intelligence Lab (MagCIL). The package aims to simplify the feature extraction and classification process by providing a number of helpful tools at can sift through the signal and create relevant features. These features can then be used to train models for classification tasks.

This might be relevant, which has a set of feature extraction methods implemented: https://github.com/tyiannak/pyAudioAnalysis/wiki/3.-Feature-Extraction

A library for audio feature extraction, regression, classification, segmentation\ (2 comments)

This doesn't seem hard, you can use something like this to dwoanload the songs: https://stackoverflow.com/a/27481870/6151784 and something like this to calculate how much they match: https://github.com/worldveil/dejavu The question is would you create a (dedicated) server to do your work? Or your own pc? You could also create a very simple page where someone would paste you a YouTube profile URL and you would check all songs of this URL. Also to have a db and save information about the matching and which youtube profiles have alsready been checked. Something like that could work.

If you read the installation.md file it clearly states that it has only been tested on UNIX systems, so you might be on your own trying to get it to wor in windows.

Using neural networks seems complicated, did you tried audio fingerprinting? I have been using this audio fingerprinting library to power this anime song synchronization script. You can check Panako and dejavu too.

I want to say these few seconds match these few seconds from a different audio track. Using dejavu raw has overhead I do not need/want and hence I've been fiddling around with the fingerprint script. When modifying the global variables I can get better hits or worse hits, I will admit that even after reading there recommended article and many other sources, I can't find some good explanations about the mathematics behind the filtering after the specgram has been applied. As far as a I am aware we first apply filters to find/make fine points across the spectrogram after that we only check the distance between points along the time axis not the frequency or a hypotenuse (weird).

I used a repo called https://github.com/worldveil/dejavu to compare audio hashed fingerprints and distinguish the difference between them.