Wednesday, April 22, 15:00 - 17:30, Location: Show and Tell Area A
Francesco Nesta, Piergiorgio Svaizer, Maurizio Omologo
In this demo we want show the effectiveness of the recently developped real-time BSS prototype and already presented in the workshop HSCMA2008 and IWAENC2008. The system is able to separate in real-time two moving sources by using a standard laptop, even in higly reverberant environment. Two loudspeakers will play some audio files (2 speeches or 1 speech + 1 song) and the sound will be recorded in real-time by two microphones. By a switch selector and a common headphone the audience will be able to select and listen each separated source. The system is able to process independetly very short block of signals (0.5-1s) and for this reason is able to track the change in the mixing conditions when the sources (or the microphones) move.
A real-time graphical representation of the estimated demixing filters and of the estimated TDOAs will also help understanding of the algorithm.
By using the same engine of the real-time BSS algorithm another demonstration will show the attractive property of the cSCT. Will be shown that by using only two microphones it is possible to estimate in real-time the TDOAs of many speakers which activities overlap in time.