Bahador Makki Abadi, 
Department of Computing Faculty of  Engineering and Physical Sciences
University of Surrey
Email: B.Makkiabadi@Surrey.ac.uk


These results are achieved by my tensor factorization based method
which decomposes the tensor data provided by segmented mixture
signals. This work can be considered as an extension of my previous
work published in Asilomar Conf. as: 

B. Makkiabadi and S. Sanei and D. Marshall, "A k-subspace based tensor
factorization approach for under-determined blind identification ",
Proc. of Signals, Systems and Computers (ASILOMAR) Conf., Pacific
Grove, USA, 2010

In this work no frequency transformation function (e.g. FFT or
Wavelet) has been used and the separation has been done in time domain
only.

In terms of CPU time, My system has an Intel i3 CPU running at 2.1GHz.
My codes were written for Matlab and they are not optimized for speed of running. 
Roughly measured time to recover the sources:
For 2x3 cases it takes about 1:10 minutes:seconds
For 2x4 cases about 2:00 minutes:seconds
For 3x4 cases about 2:30 minutes:seconds