1. Benedikt Loesch and Bin Yang University of Stuttgart Chair of System Theory and Signal Processing International Workshop on Acoustic Echo and Noise Control, 2008 Presenter Chia-Cheng Chen 1

2.  Introduction  Observation Vector Clustering  Source Number Estimation  Experimental results  Conclusion 2

3.  The task of blind source separation is to separate M (possibly) convolutive mixtures x m [i],m = 1,...,M into N different source signals.  Present an algorithm call NOSET (Number of Source Estimation Technique) 3

4.  Short Time Fourier transform (STFT)  Three steps ◦ Normalization ◦ Clustering ◦ Reconstruction 4

5.  Normalization ◦ The normalization is performed with respect to a reference sensor J [4] ◦ Unit-norm normalization 5

6.  Clustering ◦ K-means  Reconstruction [4] 6

7.  The phase difference among different sensors is large enough. In the low-frequency region, this is not the case and the phase estimate is rather noisy.  Only one source is dominant at a TF point [k, l]. 7

8.  Selection of One-Source TF Points  Power of source n  Selection of reliable TF points 8

9.  DOA Estimation ◦ time delay δ m for sensor m 9

10. 10

11. 11

12. 12

13.  Frequency of fs = 8 kHz and a cross-array with M = 5 microphones  16 sets of 6 speech signals (3 male, 3 female, different for each of the 16 sets)  SNR was between 20 and 30 dB  Typical values are: f l = 250Hz, t2 = 20 dB, t3 = 0.2 13

14. 14

15. 15

16.  Presented the NOSET algorithm to estimate the number of sources in blind source separation.  It relies on DOA estimation at selected one-source TF points and works in both overdetermined and underdetermined situations. 16