Audio Note Analysis

Converts a monophonic wav file into a MIDI file. A more in depth explanation can be found here

Requirements

To run, the following must be installed

• Arrayfire Binaries ( Last tested with 3.6.2 )
• CUDA Toolkit
  • If CUDA is not installed it should fallback to OpenCL or CPU
• Python3
  • Required python packages:

pip install --user numpy PyQt5 pyqtgraph matplotlib

• Make sure that the CUDA and Arrayfire Environment Variables are set correctly

Build

First build the Rust source ^[1] as a dynamic library to create wav2midi.dll with the following:

$ cargo +nightly build

1. Dependency synthrs requires nightly toolchain:

$ rustup install nightly

Usage

$ analyze.py [-h] -f FILE_NAME [-w WINDOW] [-p HIGHPASS] [-r HPS_RATE]
          [-o ONSET_WINDOW] [-c THRESHOLD_CONSTANT] [-i] [-n]

OR

interface.py provides a GUI

Standalone (CLI only)

$ make

will create a standalone distributable folder if Rust Toolchain and Python3 are installed. SINGLE_FILE=1 flag will create a single executable file instead of a folder, but the execution will be slower as everytime files are unpacked into a temporary folder

Sample output example

• testaudio/tetris_acoustic_guitar.wav → testaudio/tetris_acoustic_guitar.mid

analyze.py -f testaudio/tetris_acoustic_guitar.wav -w 13 -p 215 -r 3 -o 7 -c 1.03

• testaudio/tetris_violin.wav → testaudio/tetris_violin.mid

analyze.py -f testaudio/tetris_violin.wav -w 13 -p 215 -r 6 -o 7 -c 1.02