International Audio Laboratories Erlangen

New Insights on Target Speaker Extraction

M. Elminshawi, W. Mack, S. Chakrabarty, and E. A. P. Habets

1. Performance for Different Gender Combinations

GRID - MM

Sample 0

Sample 1

Sample 2

GRID - FF

Sample 0

Sample 1

Sample 2

GRID - MF

Sample 0

Sample 1

Sample 2

TCD-TIMIT - MM

Sample 0

Sample 1

Sample 2

TCD-TIMIT - FF

Sample 0

Sample 1

Sample 2

TCD-TIMIT - MF

Sample 0

Sample 1

Sample 2

LRS3 - MM

Sample 0

Sample 1

Sample 2

LRS3 - FF

Sample 0

Sample 1

Sample 2

LRS3 - MF

Sample 0

Sample 1

Sample 2

WSJ0 - MM

Sample 0

Sample 1

Sample 2

WSJ0 - FF

Sample 0

Sample 1

Sample 2

WSJ0 - MF

Sample 0

Sample 1

Sample 2

LibriSpeech - MM

Sample 0

Sample 1

Sample 2

LibriSpeech - FF

Sample 0

Sample 1

Sample 2

LibriSpeech - MF

Sample 0

Sample 1

Sample 2