The influence of the rendering architecture on the subjective performance of blind source separation algorithms

Blind Source Separation algorithms often include a time/frequency (t/f) decomposition / filterbank as an important part allowing for frequency selective separation of the input signal. In order to investigate the importance of the t/f processing architecture for the achieved subjective audio quality, a set of blindly separated audio signals were taken from the Stereo Audio Source Separation Campaign (SASSEC) 2007 and rated in a MUSHRA listening test. The test furthermore included material that was generated by using the separated signals to drive an enhanced time/frequency rendering architecture, as it is offered by MPEG Spatial Audio Object Coding (SAOC). In this way, the same basic separation core algorithm was applied together with different t/f rendering architectures. The listening test reveals an improved subjective quality for the SAOC post-processed items.