wavelet-denoiser

wavelet-denoiser copied to clipboard

• TL;DR #+BEGIN_SRC sh python3 src/denoiser-argument.py -i -o

  #+END_SRC

  For an extensive list of what arguments you can pass #+BEGIN_SRC sh :results pp :exports both :eval never-export python3 src/denoiser-argument.py --help #+END_SRC

  #+RESULTS: #+begin_example usage: denoiser-argument.py [-h] -i INPUT [-a A] [-b B] [-c C] [-d D] [-type {1,2,3}] [-akg AKG] [-ako AKO] [-aks {asc,desc}] [-l L] [-wavelet WAVELET] [-method {wpa,dwt}] [-t TIME] [-v] -o OUTPUT

  optional arguments: -h, --help show this help message and exit -i INPUT, --input INPUT the relative or absolute path of the sound file you wish to denoise -a A denoiser param 'a' (default: 2) -b B denoiser param 'b' (default: 1) -c C denoiser param 'c' (default: 1) -d D denoiser param 'd' (default: 0.1) -type {1,2,3} filter type (default: 1) -akg AKG grad of the Ak filter (default: 4) -ako AKO offset of the Ak filter (default: 2) -aks {asc,desc} the slope of the Ak filter - 'asc' or 'desc' (default: asc) -l L wavelet packed decomposition levels (default: 8) -wavelet WAVELET the wavelet to be used (default: db8) -method {wpa,dwt} The wavelet transform method - 'wpa' or 'dwt' (default: wpa) -t TIME, --time TIME the period of silence present in the audio file (in seconds) eg: '0-0.5'. If none provided, the noise period will be autoimatically found -v, --verbose verbose (plays the result audio) -o OUTPUT, --output OUTPUT The output filename #+end_example comment: =org-mode= in =emacs= is awesome for writing this kind of documents :)

** Docker way First run =docker-compose build=, and then

#+BEGIN_SRC sh :results pp :exports both :eval never-export docker-compose run denoiser python3 src/denoiser-argument.py -i test/resources/sp01_airport_sn5_44100.wav -o denoised.wav #+END_SRC

You should see something like #+RESULTS: #+begin_example trying to open test/resources/sp01_airport_sn5_44100.wav Number of samples read: 124192 Denoiser options: a: 2 b: 1 c: 1 d: 0.1 akGrad: 4 akOffset: 2 filterType: 1 method: wpa wavelet: db8 Noise profiler finished 0% 1% ... 99% 100% will write denoised file to denoised.wav OK #+end_example

To denoise your own file (let's say =noised.wav=)

• put this file in the root directory
• run =docker-compose run denoiser python3 src/denoiser-argument.py -i noised.wav -o denoised.wav=
• profit :)

If you don't want to use =docker-compose= and only docker

• =docker build . -t actondev/denoiser=
• and then #+BEGIN_SRC sh docker run -v $(pwd):/code actondev/denoiser python3 src/denoiser-argument.py -i test/resources/sp01_airport_sn5_44100.wav -o denoised.wav #+END_SRC

• Wavelet denoiser This is the product of my thesis for my =Electrical & Computer Engineering= diploma at =Aristotle University of Thessaloniki=. Good times (nah not really. i'm lazy, and doing this while having a day-job.. mierda)

  Open sourced for 2 reasons

  • other people can use it
  • other people can improve it/comment on how to improve it ..aka tell me how much shitty the code -or the math- is :)

  It's my first attempt to try to do something in the open source world.

  #+BEGIN_SRC sh :exports both :eval never-export echo "Hello open source world" #+END_SRC

  #+RESULTS: : Hello open source world

• Installation

  • sudo apt-get install python3 python3-setuptools python3-pip
  • pip3 install -r requirements.txt or instad of =pip3= you can run =python3 -m pip=

** Windows

• [[https://chocolatey.org/install][install chocolatey]]
• choco install python3 with the above, python3 is available with the =python= command. Change this to python3 like that, by using =git bash=

** python3 vs python fix If you are on windows, and your executable to run =python= is.. =python= (even though it's version 3) then using =git bash=:

• =which python= will tell you the path of the executable for the python
• make sure =python -V= shows version 3
• =ln -s /path/of/python/from/which/python/command ~/bin=
• edit the environment variables for your current user, and add the =C:\users\USERNAME\bin= there

For example, I had to run =ln -s /c/Python36/python.exe ~/bin/python3.exe=

This wall make =python3= available in the =cmd= console of windows as well (useful in case you run the tests from =Visual Studio Code= in Windows)

A simpler way, to just make the =python3= command available in =git bash= would be to

• =vim ~/.bashrc= #+BEGIN_SRC text alias python3="python" #+END_SRC
• =source ~/.bashrc=

• Usage

** Denoiser run =python3 denoiser-argument.py -i -o

=. For all the parameters you can pass to the denoiser, run =python3 denoiser-argument --help=

** Denoised file metrics If you have the clean audio file (that is then noisified)

• denoise the file
• calculate the denoiser metric \ =python3 src/metric-cci.py -a "the clean file .wav" -b "the denoised file.wav"=

** Available wavelets ['bior1.1', 'bior1.3', 'bior1.5', 'bior2.2', 'bior2.4', 'bior2.6', 'bior2.8', 'bior3.1', 'bior3.3', 'bior3.5', 'bior3.7', 'bior3.9', 'bior4.4', 'bior5.5', 'bior6.8', 'cgau1', 'cgau2', 'cgau3', 'cgau4', 'cgau5', 'cgau6', 'cgau7', 'cgau8', 'cmor', 'coif1', 'coif2', 'coif3', 'coif4', 'coif5', 'coif6', 'coif7', 'coif8', 'coif9', 'coif10', 'coif11', 'coif12', 'coif13', 'coif14', 'coif15', 'coif16', 'coif17', 'db1', 'db2', 'db3', 'db4', 'db5', 'db6', 'db7', 'db8', 'db9', 'db10', 'db11', 'db12', 'db13', 'db14', 'db15', 'db16', 'db17', 'db18', 'db19', 'db20', 'db21', 'db22', 'db23', 'db24', 'db25', 'db26', 'db27', 'db28', 'db29', 'db30', 'db31', 'db32', 'db33', 'db34', 'db35', 'db36', 'db37', 'db38', 'dmey', 'fbsp', 'gaus1', 'gaus2', 'gaus3', 'gaus4', 'gaus5', 'gaus6', 'gaus7', 'gaus8', 'haar', 'mexh', 'morl', 'rbio1.1', 'rbio1.3', 'rbio1.5', 'rbio2.2', 'rbio2.4', 'rbio2.6', 'rbio2.8', 'rbio3.1', 'rbio3.3', 'rbio3.5', 'rbio3.7', 'rbio3.9', 'rbio4.4', 'rbio5.5', 'rbio6.8', 'shan', 'sym2', 'sym3', 'sym4', 'sym5', 'sym6', 'sym7', 'sym8', 'sym9', 'sym10', 'sym11', 'sym12', 'sym13', 'sym14', 'sym15', 'sym16', 'sym17', 'sym18', 'sym19', 'sym20']

..though some give errors. See more [[https://stackoverflow.com/questions/41189110/why-pywavelet-wavelet-does-not-accept-all-built-in-wavelets][here]]

• Running the tests #+BEGIN_SRC sh python3 -m unittest discover -v -s ./test -p 'test.py' #+END_SRC

• Continuous Integration Had the project initially just on gitlab and started using =gitlab-ci=.. seems quite nice :) It can integrate with public github repos as well. Hooray

• Known problems

  • [ ] the denoiser is optimized for input files of =44100 kHz= samplerate.