Perceiving foreign-accented speech with decreased spectral resolution in single- and multiple-talker conditions

Abstract

To determine the effect of reduced spectral resolution on the intelligibility of foreign-accented speech, vocoder-processed sentences from native and Mandarin-accented English talkers were presented to listeners in single- and multiple-talker conditions. Reduced spectral resolution had little effect on native speech but lowered performance for foreign-accented speech, with a further decrease in multiple-talker conditions. Following the initial exposure, foreign-accented speech with reduced spectral resolution was less intelligible than unprocessed speech in both single- and multiple-talker conditions. Intelligibility improved with extended exposure, but only for single-talker conditions. Results indicate a perceptual impairment when perceiving foreign-accented speech with reduced spectral resolution.

The Perception of Regional Dialects and Foreign Accents by Cochlear Implant Users

“…CI users have also been found to have more difficulty recognizing highly variable speech produced by multiple talkers from different dialect regions (e.g., Sommers et al, 1997;Tamati et al, 2020) and foreignaccented talkers (Ji et al, 2014;Kapolowicz et al, 2020), compared to idealized lab speech, which typically involves carefully articulated speech produced by a single talker with no discernable accent. Similarly, studies using simulations of CI hearing have also reported less accurate recognition of highly variable speech (Faulkner et al, 2015) and foreignaccented speech (Kapolowicz et al, 2016(Kapolowicz et al, , 2018. Thus, although CIs improve hearing, different sources of indexical variability may pose substantial problems to speech recognition.…”

Effects of Spectral Envelope and Fundamental Frequency Shifts on the Perception of Foreign-Accented Speech

“…Presentation of these five talkers was randomly interspersed for each block, and listeners heard each talker twice per block (additional details regarding blocks are described below). Data from both unshifted foreign-accented conditions were reported in our previous work (Kapolowicz et al, 2018), and again, presented here to allow for direct comparisons between shifted and unshifted speech. Since listeners were different for each condition, and all listeners shared the same relevant demographical information, statistical comparisons across conditions are allowed; (c) for the Foreign-Accented Single Talker Shifted condition, listeners heard unshifted speech as well as spectrally shifted speech from the same foreign-accented talker (randomly selected for each listener from one of the five talkers in the Foreign-Accented Multiple Talkers Unshifted condition) across 40 sentences to simulate five different talkers.…”

“…When these cues vary as they do across different talkers, rapid adaptation is impaired. Conversely, we did not expect spectral shifts to affect intelligibility of native-accented speech, since our previous work using the same paradigm found no difference in intelligibility scores across single-talker and multipletalker conditions (Kapolowicz et al, 2018). The native-accented spectrally shifted condition was included for comparative purposes and to demonstrate that listeners have no trouble perceiving spectrally-shifted native-accented speech using the signal processing implemented in the present work, which resulted in clean speech signals, with minimal artifacts/distortions that could affect intelligibility (as tested in Experiment 2).…”

“…Increased exposure to speech from a single native-accented talker over time leads to improved recognition accuracy (Magnuson & Nusbaum, 2007). Similarly, increased exposure to foreign-accented speech from a single talker leads to improved recognition accuracy (Kapolowicz et al, 2018). This suggests that talker normalization contributes to the initial process of adapting to foreign-accented speech, just as it does in native-accented speech.…”

“…These results reveal that listeners rely on spectral pattern information to a greater extent for foreign-accented speech than for native-accented speech. A follow-up study using a 9-channel vocoder (Kapolowicz et al, 2018) further revealed that intelligibility performance when perceiving spectrally reduced native-accented speech was near ceiling when listeners heard only one talker over time as well as when listeners heard multiple interleaved talkers. Intelligibility scores were much lower when perceiving spectrally reduced foreign-accented speech, and no rapid accent-dependent adaptation (i.e., no improvement with increased exposure) occurred in conditions where listeners were exposed to multiple foreign-accented talkers despite sharing the same native language.…”