Spoken word recognition in noise in Mandarin-speaking pediatric cochlear implant users

Objective: The purpose of the present study was to compare spoken word recognition performance in the presence of speech spectrum-shaped noise and four-talker babbles in Mandarin-speaking children with cochlear implants (CIs). Methods: Participants included 33 children with unilateral CIs (with a mean age of 10.4 +/- 2.9 years old and a mean length of CI use of 7.5 +/- 3.0 years). The Standard Chinese version of Lexical Neighborhood Test was implemented in quiet, speech-spectrum-shaped noise (SSN), and four-talker babble (FTB). The signal-to-noise ratios (SNRs) were set at +5 and + 10 dB for both types of maskers. Participants responded by verbally repeating each word they heard and the response was scored as the percentage accuracy of recognition performance. A Generalized Linear Model (GLM) fitting, correlational tests, and a two-way repeated-measures ANOVA were conducted on the percent-correct data. Results: Word recognition in quiet was on average 74.5% correct but dropped to 57.3% and 48.8% correct for SSN and FTB at 10 dB SNR, respectively, and 44.4% and 32.6% correct for SSN and FTB at 5 dB SNR, respectively. In both quiet and noise conditions, the participants showed lower recognition accuracy for the hard words than for the easy words. Disyllabic words were recognized with higher accuracy rates than were the monosyllabic words. The GLM analysis revealed that all four tested factors (masker type, SNR, lexical neighborhood feature, and lexical type) showed significant impacts on word recognition in children with CIs. Word recognition scores in the two types of maskers were significantly correlated for the disyllabic words at both SNRs and monosyllabic words at 10 dB SNR. Conclusions: The present study demonstrated that the lexical features such as the lexical neighborhood characteristics and lexical type had significant effects on speech recognition performance in both quiet and noise conditions in pediatric CI users. Children with years of experience of CI use still encountered remarkable difficulties in everyday listening environment although their speech recognition in quiet reached relatively desired level. Fluctuating noise, such as speech babbles, caused greater challenge than steady-state noise for speech recognition in children with CIs.