14th Speech in Noise Workshop, 12-13 January 2023, Split, Croatia 14th Speech in Noise Workshop, 12-13 January 2023, Split, Croatia

P01Session 1 (Thursday 12 January 2023, 15:30-17:30)
Speech processing in multi-talker situations: The role of speaker similarity

Petra Kovács
Institute of Cognitive Neuroscience and Psychology, Research Centre for Natural Sciences, Budapest, Hungary
Budapest University of Technology and Economics, Budapest, Hungary

Brigitta Tóth
Institute of Cognitive Neuroscience and Psychology, Research Centre for Natural Sciences, Budapest, Hungary

Orsolya Szalárdy
Institute of Cognitive Neuroscience and Psychology, Research Centre for Natural Sciences, Budapest, Hungary
Semmelweis University, Budapest, Hungary

István Winkler
Institute of Cognitive Neuroscience and Psychology, Research Centre for Natural Sciences, Budapest, Hungary

Speakers with perceptually similar voice should be harder to segregate in multi-talker situations. However, while the dissimilarity of voices may help their segregation, the same variable may also affect sustained selective attention. Thus, the overall effect of voice similarity on following one voice in the cocktail party situation may either be beneficial or detrimental. To tease the two effects apart, we collected electrophysiological (EEG) and behavioral data while 22 healthy young adults listened to two concurrent speech streams consisting of either 1) identical, 2) similar, 3) dissimilar, or 4) opposite-gender speakers. Functional brain connectivity and behavioral results suggested that, while speaker similarity hinders auditory stream segregation, dissimilarity hinders selective attention by making the speech stream to be ignored more distracting. The problem to be solved by the speech processing system in multi-talker situations is thus different depending on the level of perceived speaker similarity.

Last modified 2023-01-06 23:41:06