How to extract musical features from an audio fragment?

Question

Machine learning

What existing tools can extract the maximum musical information from audio? Assume the tempo, tonality, determine a solo instrument or several voices at the same time, with accompaniment, or is there just background noise. Try to classify the genre. Assume that it is rather the drums that are portrayed by the mouth, or the melody is whistled lispingly.

The idea is to find from the whistlings sent by so-so quality to the microphone of the phone those close in “spirit” from the base of works (which also have to be indexed in this way).

Answer 1

Assume the tempo, tonality, determine a solo instrument or several voices at the same time, with accompaniment or is there just background noise.

First thought: look on the Internet.
As for whistling, I doubt very much about the idea.
And here

Assume that it is rather the drums that are portrayed by the mouth, or the melody is whistled lispingly.

closer to fantasy today.
Try to start relatively simple.