Bracha Laufer-Goldshtein, Ronen Talmon and Sharon Gannot, "Source counting and separation based on simplex analysis," submitted to IEEE Transactions on Signal Processing, 2018.

In our experimental study, we consider both simulated data and real-life recordings:

1. Simulation: Clean speech signals (from the TIMIT database) were convolved with room impulse responses from the RIR database [1] recorded in the acoustic lab at Bar-Ilan University, with reverberation time set to 360ms.
2. Real-life Experiment: Real recordings of real-life speakers were carried out in a low echoic noisy enclosure as described in [2].

Simulation:

 

J	Mixture	Speaker 1	Speaker 2	Speaker 3	Speaker 4
2
2
2
2
2
2
3
3
3
4
4
4

 

Real-life experiment (Speaker 1 louder than Speaker 2 by 15dB):

J	Mixture	Speaker 1	Speaker 2
2
2
2

 

1. E. Hadad, F. Heese, P. Vary, and S. Gannot, “Multichannel audio database in various acoustic environments,” in International Workshop on Acoustic Signal Enhancement (IWAENC), 2014, pp. 313–317.
2. S. E. Chazan, J. Goldberger, and S. Gannot, “DNN-based concurrent speakers detector and its application to speaker extraction with LCMV beamforming,” in IEEE International Conference on Audio and Acoustic Signal Processing (ICASSP), 2018.