Unmixing Acoustic Sources in Real Reverberant Environments for Close-Microphone Applications*
Elias K. Kokkinis, John Mourjopoulos · Journal of the Audio Engineering Society · 2010
The problem of acoustic source separation is addressed in an audio application context, especially when considering close-microphone applications and the effects of room acoustics. The framework of blind source separation (BSS) is used here to formally describe the problem, and solutions derived from this framework are examined for the case of separating two interfering acoustic sources in various realistic rooms and conditions. Specifically the performance of a benchmark convolutive BSS algorithm and a Wiener filter, considered in this context as a special case of BSS algorithm, is compared via objective measures as well as a subjective evaluation test to assess their performance. The Wiener filter was found to achieve superior performance and proved to be a simple, efficient method to separate two audio sources in such applications.