This repository contains the thought process and code for understanding music.
- Taking one track, psytrance from Astrix from soundcloud in mp3 format.
- using Librosa to understand what this track is made of in an analog way.
- Once we understand the signal composition, will identify what are the different signals using bands, freq, wavelength and try to understand at every 3-5 second windows, how are they combined.
- Will start building a vector for every window based on the composition of the signals, want to call them mini-music-vectors.
- Using mini-music-vectors, will combine them to form a vector for the entire song.
- Then will run this for all the tracks from 3 artisits.
- Wil try and group these vectors to see if similar tracks are closer
- Then will try to verify at different time windows, what the vectors look like .
- Eventually, it can do what shazam did, however that is not what I am aiming for. I am thinking about it like a creative assistant to psytrance producers, where after they make the initial composition and need some creative inspiration on how to progress with their tracks, they can upload their created part and find vectors of songs that are similar to what they did and how they progressed in the next time windows.