as stated in the title of the question - What technologies I may use to write drum-pattern audio signal based recognition program? I want to create a tool for me as a drummer musician to transcribe a drum-part from a record. I imagine this as similiar technology to speech-recog but made especially for drum patterns previously defined in some kind of a drum pattern base. The problem is Im a very beginner in programming. For half year i was interested in microcontrollers with basic c++, not even OOP. Currently im trying out python and this is my final programming knowledge/experience background. Now because of that poor level of know-how in IT, I dont really know what technology, frameworks etc I should get interested in with this kind of a project. It may be obvious that I should look for speech-recognition technologies and learn that and then apply that knowledge to build my own program, but Im not really sure where is the best place to start and if im ready for reading heavy walls of professsional open source projects code. Maybe there is some kind of a friendly python framework to get me started in that topic? I found Python Librosa framework in my research but it seems really advanced and it looks like i should learn signal-theory to get fluent in using that. Let me know what do you think and what kind of tactic should i aim to in your opinion.
1 Answers
The task of transcribing music automatically, from audio into notes (typically MIDI), is known in the research community as Automatic Music Transcription. The specialized task of doing it on drums only is known as Automatic Drum Transcription (ADT).
ADT is widely researched, and both open-source and commercial solutions are available. One open-source software can be found in ADTLib. It provides a very simple Python API that takes a WAV file and returns transcribed drum track. There are papers linked in the README file, describing how it is put together. A web-based tool called ADTWeb allows to try out transcribing drums without installing any software.
Note that ADT usually assumes an input that is only/predominantly drums. If you want to extract drum patterns from a mixed song containing also other instruments, you mayb need some kind of Source Separation step as well.

- 5,494
- 1
- 21
- 50
-
2For state of the art drum transcription, check out [DrummerNet](https://github.com/keunwoochoi/DrummerNet). – Hendrik Nov 16 '20 at 13:04