Recognition of noise-vocoded accented speech by non-native listeners

Abstract

Abstract This study examined 1) how spectral degradation and talker accent strength interact to influence non-native speech perception; and 2) how sentence context contributes to non-native listeners’ perception of spectrally degraded foreign-accented speech. Twenty-eight Mandarin-speaking English-learning adults were tested on the recognition of English standard hearing test sentences and high- and low-predictability sentences produced by four Mandarin-accented English talkers (two mildly and two moderately accented) in unprocessed and 6-channel noise-vocoded conditions. Results showed poorer recognition in the vocoded than in the unprocessed conditions and for moderately accented than for mildly accented talkers. While increased accent had a limited negative effect in the optimal conditions (e.g., unprocessed speech with sentence context), it posed substantial difficulties under the suboptimal conditions (e.g., spectrally degraded speech with or without sentence context). Sentence context facilitated the recognition accuracy of moderately accented speech in the unprocessed condition but provided reduced benefits in the vocoded condition for both accent levels. Non-native listeners’ vocabulary knowledge was associated with their perceptual performance, with a stronger influence in the unprocessed than in the vocoded conditions, suggesting that spectral degradation constrained listeners’ ability to utilize lexical knowledge during speech recognition.