Page 41 - Spring2019

served (e.g., Wightman and Kistler, 2005; Nishi et al., 2010). The pronounced difficulties experienced by younger school- age children are somewhat perplexing in light of data indi- cating that the peripheral encoding of sound matures early in life. It has been posited that these age effects reflect an im- mature ability to recognize degraded speech (e.g., Eisenberg et al., 2000; Buss et al., 2017). It has also been suggested that children’s immature working memory skills also play a role in their speech-in-noise difficulties (McCreery et al., 2017).
Maturation of Auditory Scene Analysis
Children are at a disadvantage relative to adults when lis- tening to speech in competing noise, but the child/adult difference is considerably larger when the maskers are also composed of speech. Hall et al. (2002) compared word rec- ognition for children (5-10 years) and adults tested in each of two maskers: noise filtered to have the same power spec- trum as speech (speech-shaped noise; see Multimedia File 1 at acousticstoday.org/leibold-media) and competing speech composed of two people talking at the same time (see Multi- media File 2 at acousticstoday.org/leibold-media). On aver- age, children required a 3 dB more favorable SNR relative to adults to achieve a comparable performance in the noise masker. This disadvantage increased to 8 dB in the two-talker masker. In addition to the relatively large child/adult differ- ences observed in the two-talker masker relative to the noise masker, the ability to recognize masked speech develops at different rates for these two types of maskers (e.g., Corbin et al. 2016). Although adult-like speech recognition in compet- ing noise emerges by 9-10 years of age (e.g., Wightman and Kistler, 2005; Nishi et al., 2010), speech recognition perfor- mance in a two-talker speech masker is not adult-like until 13-14 years of age (Corbin et al., 2016). This prolonged time course of development appears to be at least partly due to immature sound segregation and selective attention skills. Recognition of speech produced by the teacher is likely to be limited more by speech produced by other children in the classroom than by noise produced by the projector (see Fig- ure 1). The term informational masking is often used to refer to this phenomenon (e.g., Brungart, 2005).
An important goal for researchers who study auditory de- velopment is to characterize the factors that both facilitate and limit children’s ability to perform auditory scene analysis (e.g., Newman et al., 2015; Calandruccio et al., 2016). For lis- teners of all ages, the perceptual similarity between target and masker speech affects performance in that greater masking is associated with greater perceptual similarity. A common ap- proach to understanding the development of auditory scene
analysis is to measure the extent to which children rely on acoustic voice differences between talkers to segregate target from masker speech (e.g., Flaherty et al., 2018; Leibold et al., 2018). For example, striking effects have been found between conditions in which the target and masker speech are pro- duced by talkers that differ in sex (e.g., a female target talker and a two-male-talker masker) and conditions in which tar- get and masker speech are produced by talkers of the same sex (e.g., a male target talker and a two-male-talker masker). Dramatic improvements in speech intelligibility, as much as 20 percentage points, have been reported in the literature for sex-mismatched relative to sex-matched conditions (e.g., Helfer and Freyman, 2008).
School-age (Wightman and Kistler, 2005; Leibold et al., 2018) and 30-month-old (Newman and Morini, 2017) children also show a robust benefit of a target/masker sex mismatch, but infants younger than 16 months of age do not (Newman and Morini, 2017; Leibold et al., 2018). Leibold et al. (2018), for example, measured speech detection in a two-talker masker in 7- to 13-month-old infants and in adults. Adults performed better when the target word and masker speech were mismatched in sex than when they were matched. In sharp contrast, infants performed similarly in sex-matched and sex-mismatched conditions. The overall pattern of re- sults observed across studies suggest that the ability to take advantage of acoustic voice differences between male and female talkers requires experience with different talkers be- fore the ability emerges sometime between infancy and the preschool years.
Although children as young as 30 months of age benefit from a target/masker sex mismatch, the ability to use more subtle and/or less redundant acoustic voice differences may take longer to develop. Flaherty et al. (2018) tested this hypoth- esis by examining whether children (5-15 years) and adults benefited from a difference in voice pitch (i.e., fundamen- tal frequency; F0) between target words and a two-talker speech masker, holding other voice characteristics constant. As previously observed for adults (e.g., Darwin et al, 2003), adults and children older than 13 years of age performed substantially better when the target and masker speech dif- fered in F0 than when the F0 of the target and masker speech was matched. This improvement was observed even for the smallest target/masker F0 difference of three semitones. In sharp contrast, younger children (<7 years) did not benefit from even the most extreme F0 difference of nine semitones. Moreover, although 8-12 year olds benefitted from the largest F0 difference, they generally failed to take advantage of more
Spring 2019 | Acoustics Today | 39