Unwanted SoUnd RePlacement
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
docs
noise_replacement
recurse
samples
sound-resynthesis
utilities
.gitignore
LICENSE
README.rst
__init__.py

README.rst

USURP (Unwanted SoUnd RePlacement)

The Child Mind Institute is developing USURP (Unwanted SoUnd RePlacement) as an open source software package to automate the replacement of unwanted sounds with sounds comparable to ambient noise in the rest of a given soundfile.

Release:|release|
Date:|today|

Links:

.. toctree::
    :maxdepth: 1

    license

.. todo:: `sound resynthesis <https://github.com/ChildMindInstitute/USURP/sound_resynthesis>`_

Inputs

  • audio file : .wav (16-bit)
  • labels : .aud or .xml

audio file

USURP takes 16-bit waveform soundfiles (.wav) as inputs. USURP is capable of converting the following filetypes to 16-bit waveforms:

  • Material eXchange Format (.MXF or .mxf)
  • MPEG Audio Layer III (.mp3)
  • 32-bit waveform (.wav)

labels

Currently, USURP also needs start times and stop times marked in a labeled XML file, which can be generated in Audacity. Unwanted sounds must be silenced and have their temporal boundaries marked for optimal performance.

Processing

  • If necessary, convert inputs to 16-bit waveforms.
  • Clone
    • Find a likely ambient clip elsewhere in the file by choosing a segment with no peaks at or above a given threshold (default threshold: median of the amplitude envelope of the entire file). Copy that clip into each of the segments labeled for replacement.
  • Timeshift
    • Collapse the marked silences, shortening the file.
  • Output the new files.

Outputs

  • Clone replacement (.wav)
  • Timeshifted replacement (.wav)