Underdetermined source separation using Directional Laplacian Mixture Models

Some samples demonstrating the algorithms and the experiments proposed and conducted in the following paper are presented here:



Two-Sensor Case

Groove
Sources: Source l - Source 2 - Source 3 - Source 4
Mixture: Mixed Signal (Stereo)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3 - Source 4
MDLD: Source 1 - Source 2 - Source 3 - Source 4
GaussSep: Source 1 - Source 2 - Source 3 - Source 4
MoWL: Source 1 - Source 2 - Source 3 - Source 4
Hyvarinen: Source 1 - Source 2 - Source 3 - Source 4



Latino 1
Sources: Source l - Source 2 - Source 3 - Source 4
Mixture: Mixed Signal (Stereo)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3 - Source 4
MDLD: Source 1 - Source 2 - Source 3 - Source 4
GaussSep: Source 1 - Source 2 - Source 3 - Source 4
MoWL: Source 1 - Source 2 - Source 3 - Source 4
Hyvarinen: Source 1 - Source 2 - Source 3 - Source 4


Latino 2
Sources: Source l - Source 2 - Source 3 - Source 4
Mixture: Mixed Signal (Stereo)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3 - Source 4
MDLD: Source 1 - Source 2 - Source 3 - Source 4
GaussSep: Source 1 - Source 2 - Source 3 - Source 4
MoWL: Source 1 - Source 2 - Source 3 - Source 4
Hyvarinen: Source 1 - Source 2 - Source 3 - Source 4

Dev2 Male Speech Dataset 4
Sources: Source l - Source 2 - Source 3 - Source 4
Mixture: Mixed Signal (Stereo)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3 - Source 4
MDLD: Source 1 - Source 2 - Source 3 - Source 4
GaussSep: Source 1 - Source 2 - Source 3 - Source 4
MoWL: Source 1 - Source 2 - Source 3 - Source 4
Hyvarinen: Source 1 - Source 2 - Source 3 - Source 4

Dev2 Female Speech Dataset 4
Sources: Source l - Source 2 - Source 3 - Source 4
Mixture: Mixed Signal (Stereo)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3 - Source 4
MDLD: Source 1 - Source 2 - Source 3 - Source 4
GaussSep: Source 1 - Source 2 - Source 3 - Source 4
MoWL: Source 1 - Source 2 - Source 3 - Source 4
Hyvarinen: Source 1 - Source 2 - Source 3 - Source 4

Dev2_WDrums
Sources: Source l - Source 2 - Source 3
Mixture: Mixed Signal (Stereo)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3
MDLD: Source 1 - Source 2 - Source 3
GaussSep: Source 1 - Source 2 - Source 3
MoWL: Source 1 - Source 2 - Source 3
Hyvarinen: Source 1 - Source 2 - Source 3

Dev1_WDrums
Sources: Source l - Source 2 - Source 3
Mixture: Mixed Signal (Stereo)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3
MDLD: Source 1 - Source 2 - Source 3
GaussSep: Source 1 - Source 2 - Source 3
MoWL: Source 1 - Source 2 - Source 3
Hyvarinen: Source 1 - Source 2 - Source 3

Three-Sensor Case

Dev3 Female Speech Dataset
Sources: Source l - Source 2 - Source 3 - Source 4
Mixture: Mixed Signal (three channels)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3 - Source 4
MDLD: Source 1 - Source 2 - Source 3 - Source 4
GaussSep: Source 1 - Source 2 - Source 3 - Source 4


Example 3x5
Sources: Source l - Source 2 - Source 3 - Source 4 - Source 5
Mixture: Mixed Signal (three channels)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3 - Source 4 - Source 5
MDLD: Source 1 - Source 2 - Source 3 - Source 4 - Source 5
GaussSep: Source 1 - Source 2 - Source 3 - Source 4 - Source 5

Four-Sensor Case

Example 4x8
Sources: Source l - Source 2 - Source 3 - Source 4 - Source 5 - Source 6 - Source 7 - Source 8
Mixture: Mixed Signal (four channel)
Separated Sources:
WMDLD: Source 1 - Source 2 - Source 3 - Source 4 - Source 5 - Source 6 - Source 7 - Source 8
MDLD: Source 1 - Source 2 - Source 3 - Source 4 - Source 5 - Source 6 - Source 7 - Source 8
GaussSep: Source 1 - Source 2 - Source 3 - Source 4 - Source 5 - Source 6 - Source 7 - Source 8