Individual Differences in Auditory Feedback Control of Speech in Noise

Abstract— The study tested the hypothesis that speech changes in noise (the Lombard effect) may have significant individual differences, including those due to the gender and age of the speakers. For this, the characteristics of Lombard speech were studied for 12 speakers (6 men, 6 women; age groups: 25–35 and 55–59 age). The study used recordings of speech consisting of disyllabic words with stressed vowel sounds [a], [i], [u] of Russian speech in silence and in multitalker noise at levels of 60 and 72 dB(A). Changes in the fundamental frequency (ΔF0) and intensity (ΔI) of the voice in noise compared to silence were determined. When comparing groups of men and women, significant differences in the change of F0 in noise of 60 dB are shown. Differences in vowel characteristics between the young and middle-aged speaker groups were found for ΔF0 and ΔI in 72 dB noise. Regardless of gender and age, two types of speakers were identified, differing in the ΔF0 and ΔI values at both noise levels. Type 1 speakers in multitalker noise increased F0 by 23 and 57 Hz, for levels of 60 and 72 dB, respectively, and type 2 speakers, by 16 and 23 Hz. At two levels of noise masking, the voice intensity of type 1 speakers increased by 8 and 16 dB; the voice intensity of type 2 speakers, by 6 and 10 dB. The obtained differences may be determined by the greater influence of voluntary control, with an increase in the noise level in type 2 speakers we have identified.