Perceptual evaluation of a coherence suppressing down-mix method

In many application scenarios where a given number of input channels have to be mixed into a fewer number of output channels, comb-filter-free down-mixing is needed. Comb-filter artifacts originate from a superposition of highly correlated but phase shifted signal parts. To mitigate those artifacts active post-processing or temporal alignment prior to the mixing can, for instance, be applied. Recently, a new stereo-to-mono downmix method using a different paradigm has been proposed, where coherent signal parts are identified and suppressed in one channel prior to mixing. This paper reviews this method and evaluates its performance perceptually compared to previous methods with a listening test.