c@0: %% Multichannel audio source separation evaluation framework (MASSEF) c@0: % c@0: %% Overview c@0: % c@0: % The multichannel audio source separation evaluation framework is designed c@0: % to faciliate the development and evaluation of audio source separation c@0: % algorithms. The framework generates the mixture(s), calculates the ideal c@0: % binary and ratio masks, provides the mixture(s) to the separation c@0: % algorithm(s), and evaluates the outputs of the separation algorithm(s). c@0: % The framework also evaluates the ideal masks for the purposes of c@0: % comparison. c@0: % c@0: % Sources may have any number of channels; the framework evaluates each c@0: % channel. The use of iosr.bss.mixture objects facilitate the evaluation of c@0: % spatialised mixtures (e.g. binaural). c@0: % c@0: % The framework comprises two main classes: c@0: % c@0: % * —multichannel audio source separation c@0: % evaluation framework class. c@0: % * —the results class used by c@0: % MASSEF. c@0: % c@0: % c@0: % c@0: % MASSEF can be used to evaluate and compare binaural separation algorithms, c@0: % provided that the algorithms conform to the . c@0: % c@0: %% Installation c@0: % c@0: % Before using the framework, you need to download and install its c@0: % dependencies. This can be done automatically using the c@0: % method. Before each session, c@0: % MASSEF needs to be started using the method. c@0: % c@0: %% Usage c@0: % c@0: % The framework can be run in two ways: c@0: % c@0: % # by providing |iosr.bss.mixture| objects and separation algorithms, or c@0: % # by providing estimate and true source wav files. c@0: % c@0: % If 1), the framework generates the mixture(s), calculates the ideal c@0: % binary and ratio masks, provides the mixture(s) to the separation c@0: % algorithm(s), and evaluates the outputs of the separation algorithm(s). c@0: % The framework also evaluates: the ideal masks for the purposes of c@0: % comparison, and any azimuth/elevation estimates returned by the c@0: % algorithm. Use the method to operate c@0: % in this mode. c@0: % c@0: % If 2), the framework evaluates only the supplied estimate(s) using c@0: % signal-related metrics. Use the c@0: % method to operate in this mode. c@0: % c@0: % c@0: %

c@0: %