The impact of exploiting spectro-temporal context in computational speech segregation

The experimental data from the study:

https://asa.scitation.org/doi/10.1121/1.5020273

Group 1 contains results, masks and audio from the models of the 16 GMM component segregation system
Group 2 contains results, masks and audio from the models of the 64 GMM component segregation system

There are three folders:

Audio:
The CLUE sentences that were used for the listener study

IBM = Ideal Binary Mask, UP = UnProcessed, EBM = Estimated Binary Mask. 

The IBM and UP are stored in one of the configuration folders (Front-end), that is:

Audio\Group1\Front-end\icra_01_10sec_matched\UP
Audio\Group1\Front-end\icra_01_10sec_matched\IBM
Audio\Group1\Front-end\icra_01_10sec_matched\EBM

Results:
The computed metrics for group 1 & 2 as well as Word Recognition Scores (WRSs) from the listener study

BinaryMasks:

a priori SNR masks, IBMs and EBMs from group 1 and 2.

Developed with Matlab R2016a.