[aubio-user] aubio 0.4.0~beta1

[Paul Brossier]

Hi all!

Finally, I have just tagged and published aubio-0.4.0~beta1.

This can be considered as a release candidate before releasing the final
release, 0.4.0.

Please download it and test it thoroughly!

  http://aubio.org/pub/aubio-0.4.0~beta1.tar.bz2

Here are the relevant ChangeLog entries:

  * Overdue: After more than five years of development behind the curtain,
  time has come to release a new version of aubio.

  * General: The library has been completely revised since 0.3.2. The API has
  seen a major clean up, and has been thoroughly tested. The following list of
  changes is not exhaustive.

  * Memory management: allocation and freeing of memory has been optimized in
  many ways. Several memory leaks and out of bound access have been fixed.

  * Optimization: the FFT, central to most algorithm, can now be computed
  using different optimized algorithms, depending on what is available on your
  platform (FFTW, Ooura, or vDSP). Other simple optimization tricks are
  included. Most can be deactivated by configuring the build accordingly.

  * python/: The python interface has been completely rewritten to use numpy C
  interface, making the aubio python module order of magnitudes faster than
  the previous version. Several demos and tests are included.

  * src/: source and header files are now organized in sub-directories.

  * src/io/source.h: new source readers can now use any or all of libav,
  CoreAudio, and libsndfile. This means that aubio can now easily read most
  uncompressed and compressed formats. Compiled with libav, aubio can also
  read audio from video files, and over the network.

  * src/io/sink.h: a new sink object lets you write wav files with any number
  of channels, at any samplerate, using libsndfile or CoreAudio.

  * src/onset, src/tempo/, src/pitch: the different methods for onset, tempo,
  and pitch extraction have seen many bug-fixes and optimizations.

  * src/spectral/specdesc.h: new onset distances and statistical measures have
  been added.

  * src/spectral/filterbank.h: new filter bank to compute the energy in any
  custom-defined frequency bands.

  * src/spectral/mfcc.h, examples/aubiomfcc.c: a standard implementation of
  the Mel-Frequency Cepstrum Coefficients algorithm has been added.

  * src/temporal/{a,c}_weighting.h: standard implementation of the C-weighting
  and A-weighting pre-processing filters are now provided for most commons
  sampling rates.

  * src/synth/wavetable.h, src/synth/sampler.h: provide basic ways to generate
  some sounds.

  * src/fvec.h: fvec_t, the vector object central to most aubio algorithms, is
  now single channel. This simplifies the code of each algorithm greatly.

  * src/lvec.h: lvec_t provides a double precision vector, required for some
  operations to avoid floating point overflow

  * src/fmat.h: fmat_t provides a single precision matrix, useful for
  multi-channel operations and to some algorithms such as the spectral filter
  bank.

  * examples/: several new options, including new programs, have been
  included. Refer to the documentation for details.

  * tests/: several tests and examples programs have been added. This should
  be a good place to look at to understand how to use aubio.

  * doc/web.cfg: a simplified Doxygen configuration produces a simpler html
  documentation.

  * doc/*.txt: the manpages have been rewritten for txt2man.

  * Build system: the build system has been switched from autotools/automake
  to waf. Type './waf' or see README.md for instructions on how to use waf.

The final version will follow in a few days.

Any comments and improvements welcome.

Best wishes,
Paul