Analysis of Breathiness in Contextual Vowel of Voiceless Nasals in Mizo

Pamir Gogoi, Sishir Kalita, Parismita Gogoi, Ratree Wayland, Priyankoo Sarmah, S R Mahadeva PrasannaInterspeech 2018

TL;DR

This paper investigates how voiceless nasal consonants in the Mizo language influence the "breathiness" of the following vowel. The study finds that vowels following voiceless nasals exhibit acoustic characteristics of breathiness more strongly than those following voiced nasals, suggesting a cross-linguistic interaction between nasal voicing and vowel voice quality.

What Problem Does This Paper Address?

While the interaction between nasality and other speech features like voicing and vowel height has been studied, the relationship between nasality and phonation type, specifically breathiness, remains less explored. Breathiness is a voice quality characterized by audible air leakage during phonation, often associated with a wider glottal opening. Languages like Mizo, which possess both voiced and voiceless nasal consonants, offer a unique opportunity to study how the presence or absence of vocal fold vibration in a preceding nasal consonant might affect the acoustic properties of the subsequent vowel. Existing research suggests that the wider glottal opening required for voiceless sounds might influence adjacent vowels, but this specific interaction with nasals and breathiness in vowels has not been systematically investigated.

Methods

The researchers analyzed speech samples from two native Mizo speakers. They focused on the transition region of vowels following voiced and voiceless nasal consonants. To quantify breathiness, they extracted four acoustic features from the excitation source signal:

  • H1-H2 Ratio: This measures the relative amplitude of the first two harmonics of the voice. Higher H1-H2 ratios are associated with breathy phonation due to a more symmetrical glottal opening and closing.
  • Spectral Balance (SB): This feature describes the distribution of energy across different frequency bands in the voice source spectrum.
  • Strength of Excitation (SoE): This quantifies the abruptness of glottal closure. Breathy phonation, with less abrupt closure, results in lower SoE values.
  • Waveform Peak Factor (WPF): This measures the decay characteristic of the LP residual signal around glottal closure instants, reflecting the sharpness of the glottal closure.

These features were extracted from the Linear Prediction Residual (LPR) after downsampling the speech to 8000 Hz and normalizing amplitudes. Statistical analyses, including t-tests and factorial ANOVA, were used to determine the significance of the findings.

Datasets

The study used a custom dataset of 30 monosyllabic Mizo words recorded by two female native speakers. The words featured CV and CVC structures with onset nasal consonants (/m, n, ŋ/ and their voiceless counterparts) followed by one of three vowels (/a/, /i/, /u/). The final consonants in CVC words were non-nasal obstruents. The recordings were made in a sound-attenuated booth at a sampling frequency of 44.1 kHz, 24-bit WAV format.

Key Insights and Ideas

  • Voiceless Nasals "Impart" Breathiness: The most significant finding is that vowels following voiceless nasals exhibit acoustic cues more strongly associated with breathiness compared to vowels following voiced nasals. This suggests that the aerodynamic conditions created by a voiceless nasal (e.g., wider glottal aperture) can influence the voice source characteristics of the subsequent vowel.
  • Vowel Height Matters (Sometimes): While breathiness was generally higher after voiceless nasals, the degree of this effect was influenced by vowel height. For some acoustic measures (H1H2 ratio and Spectral Balance), there was a significant interaction between nasal context (voiced/voiceless) and vowel type (/a/, /i/, /u/), indicating that the effect of the preceding nasal is not uniform across all vowels.
  • Acoustic Correlates of Breathiness: The study reinforces the established acoustic correlates of breathiness (higher H1-H2, lower SoE, potentially altered SB and WPF) and demonstrates their applicability in analyzing the subtle phonetic variations in Mizo.

Key Results

  • T-tests confirmed statistically significant differences in all four acoustic parameters (H1H2, SB, SoE, WPF) between vowels following voiced and voiceless nasals.
  • Specifically, H1H2 ratio was significantly higher, and SoE, SB, and WPF were significantly lower for vowels following voiceless nasals compared to voiced nasals.
  • Factorial ANOVA revealed a significant main effect of nasal context and vowel height on all four features.
  • The interaction between nasal context and vowel type was significant for H1H2 ratio and Spectral Balance, suggesting that the breathiness effect is modulated by vowel height. For instance, /a/ and /u/ showed more breathiness after voiceless nasals, while /i/ did not show this effect as strongly for H1H2.

Limitations and Future Work

  • Limited Speaker Data: The study was based on only two female speakers, which limits the generalizability of the findings.
  • Focus on Transition Region: The analysis focused on the transition region of the vowel. Future work could explore the entire vowel duration.
  • Further Investigation of Glottal Behavior: The authors plan to investigate the glottal opening and closing behavior more directly using estimated glottal flow waveforms.
  • Nasality and Breathiness Interaction: They also aim to explore how the addition of nasal zeros in the vocal tract spectrum (due to nasalization) might interact with breathy phonation characteristics.

Summary updated

BibTeX
@inproceedings{gogoi2018analysis,
  author = {Pamir Gogoi and Sishir Kalita and Parismita Gogoi and Ratree Wayland and Priyankoo Sarmah and S R Mahadeva Prasanna},
  title = {{Analysis of Breathiness in Contextual Vowel of Voiceless Nasals in Mizo}},
  booktitle = {Interspeech},
  year = {2018},
}
Figure