Mercurial > hg > massef

%% Multichannel audio source separation evaluation framework (MASSEF)
%
%% Overview
%
% <<../MASSEF.png>>
%
% The multichannel audio source separation evaluation framework is designed
% to faciliate the development and evaluation of audio source separation
% algorithms. The framework generates the mixture(s), calculates the ideal
% binary and ratio masks, provides the mixture(s) to the separation
% algorithm(s), and evaluates the outputs of the separation algorithm(s).
% The framework also evaluates the ideal masks for the purposes of
% comparison.
%
% Sources may have any number of channels; the framework evaluates each
% channel. The use of iosr.bss.mixture objects facilitate the evaluation of
% spatialised mixtures (e.g. binaural).
%
% The framework comprises two main classes:
%
% * <help_MASSEF.html MASSEF>&mdash;multichannel audio source separation
% evaluation framework class.
% * <help_MASSEFresults.html MASSEFresults>&mdash;the results class used by
% MASSEF.
%
% <help_Example.html View an example of the use of these classes.>
%
% MASSEF can be used to evaluate and compare binaural separation algorithms,
% provided that the algorithms conform to the <help_Separators.html
% required format>.
%
%% Installation
%
% Before using the framework, you need to download and install its
% dependencies. This can be done automatically using the
% <help_MASSEF_install.html MASSEF.install> method. Before each session,
% MASSEF needs to be started using the <help_MASSEF_start.html
% MASSEF.start> method.
%
%% Usage
%
% The framework can be run in two ways:
%
% # by providing |iosr.bss.mixture| objects and separation algorithms, or
% # by providing estimate and true source wav files.
%
% If 1), the framework generates the mixture(s), calculates the ideal
% binary and ratio masks, provides the mixture(s) to the separation
% algorithm(s), and evaluates the outputs of the separation algorithm(s).
% The framework also evaluates: the ideal masks for the purposes of
% comparison, and any azimuth/elevation estimates returned by the
% algorithm. Use the <help_MASSEF_execute.html execute> method to operate
% in this mode.
%
% If 2), the framework evaluates only the supplied estimate(s) using
% signal-related metrics. Use the <help_MASSEF_evaluate.html evaluate>
% method to operate in this mode.
%
% <html>
% <hr>
% <p>Copyright &copy; 2016 <a href="http://www.surrey.ac.uk">University of Surrey</a><br>
% <a href="http://iosr.uk">Institute of Sound Recording, University of Surrey, UK</a></p>
% </html>
author	Christopher Hummersone <c.hummersone@surrey.ac.uk>
date	Thu, 02 Mar 2017 16:18:30 +0000
parents	e34a535b9af0
children	063175bfe283