direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Effects of Spatial Audio on Speaker Recognition

LOCATION:  TEL, Auditorium 1 (20th floor), Ernst-Reuter-Platz 7, 10587 Berlin

Date/Time: 08.05.2017, 14:15-15:00

SPEAKER: Lars-Erik Riechert


For participants of multi-party audio conferences, it can be challenging to attribute what was said to the individual talkers. Spatial audio reproduction may help to overcome this issue. Our work investigates the potential improvement for speaker recognition using binaural synthesis. Here, the individual talkers are distributed to different simulated positions around the listener. In a listening-only test we evaluated the effects of binaural synthesis and additional reverberation compared to a mono-diotic representation. Four similar sounding female voices were presented sequentially either as A) binaural synthesis with four virtual positions at +-12 degrees & +-30 degrees, B) binaural synthesis with identical positions at 0 degree or C) mono-diotic. Conditions A) and B) featured optional reverberation. The results show a positive effect of the binaural synthesis (A) for speaker recognition performance compared to mono-diotic. In addition, binaural synthesis with low levels of reverberation resulted in the best speaker recognition performance compared to the anechoic settings. Nevertheless, a slight negative impact of the binaural synthesis was observed for the 0 degree condition. This indicates that the processing weakens some mono-aural cues. However, this negative effect is smaller than the potential gain for speaker recognition performance if talkers are virtually distributed around the listener.

Zusatzinformationen / Extras


Schnellnavigation zur Seite über Nummerneingabe