The perception of emotion and focus prosody with varying acoustic cues in cochlear implant simulations with varying filter slopes

Daan J. Van De Velde, Niels O. Schiller, Vincent J. Van Heuven, Claartje C. Levelt, Joost Van Ginkel, Mieke Beers, Jeroen J. Briaire, Johan H.M. Frijns

Research output: Contribution to journal/periodicalArticleScientificpeer-review

5 Citations (Scopus)

Abstract

This study aimed to find the optimal filter slope for cochlear implant simulations (vocoding) by testing the effect of a wide range of slopes on the discrimination of emotional and linguistic (focus) prosody, with varying availability of F0 and duration cues. Forty normally hearing participants judged if (non-)vocoded sentences were pronounced with happy or sad emotion, or with adjectival or nominal focus. Sentences were recorded as natural stimuli and manipulated to contain only emotion- or focus-relevant segmental duration or F0 information or both, and then noise-vocoded with 5, 20, 80, 120, and 160 dB/octave filter slopes. Performance increased with steeper slopes, but only up to 120 dB/octave, with bigger effects for emotion than for focus perception. For emotion, results with both cues most closely resembled results with F0, while for focus results with both cues most closely resembled those with duration, showing emotion perception relies primarily on F0, and focus perception on duration. This suggests that filter slopes affect focus perception less than emotion perception because for emotion, F0 is both more informative and more affected. The performance increase until extreme filter slope values suggests that much performance improvement in prosody perception is still to be gained for CI users.

Original languageEnglish
Pages (from-to)3349-3363
Number of pages15
JournalJournal of the Acoustical Society of America
Volume141
Issue number5
DOIs
Publication statusPublished - 01 May 2017

Fingerprint

Dive into the research topics of 'The perception of emotion and focus prosody with varying acoustic cues in cochlear implant simulations with varying filter slopes'. Together they form a unique fingerprint.

Cite this