Alexander Lunichkin1, Alisa Gvozdeva1, Larisa Zaitseva1, Elena Ogorodnikova2, Irina Andreeva1
1Laboratory of comparative sensory physiology, Sechenov Institute of Evolutionary Physiology and Biochemistry of Russian Academy of Sciences, Saint Petersburg, Russia
2Pavlov Institute of Physiology of Russian Academy of Sciences, Saint Petersburg, Russia

Lombard reflex improves quality of speech communication in noisy environments. The reflex modifies amplitude-frequency parameters of speech, thus, increasing its intelligibility. The aim was to evaluate the spectral changes characterizing the pronunciation of vowels in speech-like noise. Six female Russian native speakers without speech and hearing impairments aged 20-58 years took part in the study. Speech recordings were carried out in an anechoic chamber during single session. Recordings of nine words with [a], [i] and [u] vowels in different stressed positions were made in quiet and in babble-noise of 60 dB(A).  Before the start of the recording, the speaker adjusted the auditory feedback to correspond to its natural level. The speaker pronounced words addressing them to the experimenter to simulate a communicative situation. Spectral characteristics (F0, F1, and F2) were analyzed on stationary intervals of stressed vowels in tool Praat. Wilcoxon Matched Pairs Test was used to compare the individual spectral parameters of vowels in quiet and in babble-noise. It was found that in noise F0 significantly increased compared to quiet for all stressed vowels in all positions. The group-averaged changes amounted to 12%. We also observed a shift in the position of vowels in the F1-F2 two-dimensional vowel space. F1 increased in babble-noise compared to quiet in all cases: for [u] it was 12% (p<0.000001), for [a] – 8% (p<0.0001), for [i] – 14% (p<0.000001). F2 changed in different ways: for [a] it increased by 2% (p<0.01), for [i] it decreased by 1% (p<0.0001). For the vowel [u] F2 was stationary. Note that in noise, the scatter of individual parameters F1 and F2 was less pronounced than in quiet. The data indicates an improvement of articulatory planning and an increase in differences between lombard vowels.

Acknowledgements: The work was supported by State budget (themes no. AAAA-A18-118050790159-4, AAAA-A18-118013090245-6).