Hearing loss and the use of acoustic cues in phonetic categorisation of fricatives

Odette Scharenborg, Esther JanseInterspeech 2012

TL;DR

As we age, our ability to hear high-frequency sounds declines, impacting our perception of speech. This study investigates how older adults with varying degrees of hearing loss distinguish between the fricative sounds /f/ and /s/, which rely heavily on high-frequency information. The findings suggest that while hearing loss doesn't hinder the use of formant transitions for this distinction, it does lead individuals to rely more on sound intensity, indicating a shift in perceptual strategies.

What Problem Does This Paper Address?

The human auditory system is remarkably adept at processing complex acoustic signals to understand speech. However, aging is associated with a natural decline in hearing sensitivity, particularly in the higher frequency ranges. This decline can interfere with the perception of speech sounds whose distinguishing features are primarily located in these high frequencies. The fricative sounds /f/ and /s/ serve as a prime example: /s/ has significant energy concentrated around 5500 Hz, while /f/ has a flatter, more distributed spectrum. Older adults with hearing loss may struggle to access this high-frequency information, potentially leading them to adopt alternative strategies for distinguishing these sounds. This paper explores what these alternative strategies might be, specifically investigating the roles of formant transitions and sound intensity in the phonetic categorization of /f/ and /s/ by older listeners with varying degrees of hearing loss.

Methods

The researchers employed two main experimental paradigms, both involving phonetic categorization tasks with older Dutch adults (aged 60+).

  • Experiment 1: Formant Transitions: This experiment focused on the role of formant transitions, which are changes in the resonant frequencies of the vocal tract as a sound is produced. To create ambiguity, the natural fricative noise at the end of words was replaced with an 11-step continuum of sounds morphing between /f/ and /s/. Crucially, the formant transitions from the preceding vowel into this ambiguous fricative were kept intact, mimicking natural speech. Participants were asked to identify the final sound as either /f/ or /s/. The hypothesis was that if hearing loss impacts the ability to use high-frequency spectral cues, listeners might rely more on formant transitions if they are preserved.
  • Experiment 2: Intensity: This experiment investigated the role of sound intensity. Here, the final fricative in the words was natural, but its intensity was systematically varied across a range from 44 dB to 56 dB (relative to a 70 dB vowel portion). Participants again performed a categorization task, identifying the final sound as /f/ or /s/. The expectation was that listeners with greater hearing loss might rely more on intensity differences, with lower intensities being perceived as /f/ and higher intensities as /s/.

Both experiments used four Dutch minimal word pairs ending in /f/ or /s/ (e.g., "brief" vs. "bries"). Participants' hearing sensitivity was assessed using standard audiometry.

Datasets

Not discussed in the paper. The stimuli were created by the researchers using recorded Dutch words and audio manipulation software (Praat, STRAIGHT, Matlab).

Key Insights and Ideas

  • Formant Transitions Remain Robust: Despite age-related hearing loss, older adults appear to effectively use formant transitions to distinguish between /f/ and /s/, even when the fricative noise itself is ambiguous. This suggests that this acoustic cue is relatively resilient to high-frequency hearing impairment in this context.
  • Hearing Loss Shifts Cue Weighting: As hearing loss increases, listeners seem to shift their reliance from spectral cues (like those in fricative noise) towards intensity cues for distinguishing /f/ and /s/. This highlights a compensatory mechanism where the auditory system adapts to reduced sensory input.
  • Language-Specific Strategies: The paper touches upon potential cross-linguistic differences in cue usage, suggesting that the native language's phonetic inventory might influence how listeners prioritize acoustic cues. Dutch, lacking a dental fricative, might lead to different strategies compared to languages like Spanish.
  • Age vs. Hearing Loss: A key question remains whether the observed reliance on formant transitions and intensity is solely due to hearing loss or if it's a general characteristic of older Dutch speakers, irrespective of their hearing status. This points to an avenue for future research.

Key Results

  • Experiment 1 (Formant Transitions): Participants consistently gave fewer /s/-responses to /f/-final source words than to /s/-final source words, indicating they could use the source word information (likely via formant transitions) to aid categorization. Importantly, there was no significant interaction between hearing loss and the use of source word information, suggesting hearing loss did not impede the use of formant transitions. However, a general effect of hearing loss was observed, with fewer /s/-responses as hearing loss increased, indicating a general difficulty in recognizing /s/.
  • Experiment 2 (Intensity): As expected, participants gave significantly fewer /s/-responses to /f/-final words than to /s/-final words. A significant effect of hearing loss was found: increased hearing loss led to fewer /s/-responses. Crucially, with increasing hearing loss, there were more /s/-responses with increasing intensity. This indicates that listeners with greater hearing loss rely more on intensity to distinguish /f/ from /s/, with higher intensities favoring /s/ identification.

Limitations and Future Work

  • Younger Listeners Comparison: The study did not include a control group of young, normal-hearing Dutch adults to directly compare their cue usage strategies. This leaves open the question of whether the observed strategies are specific to older age and hearing loss or are general to Dutch speakers. Future work is planned to investigate this.
  • Stimulus Ambiguity: While the stimuli were designed to be ambiguous, the exact degree of ambiguity and its interaction with individual listener perception could be further explored.
  • Generalizability: The findings are based on Dutch listeners and the specific /f/-/s/ contrast. Further research could explore other phonetic contrasts and languages to understand the broader implications of these findings.
  • Mechanism of Change: The paper raises the question of whether the observed shift in cue weighting is a direct consequence of hearing loss or a learned adaptation. Future experiments are needed to disentangle these possibilities.

Summary updated

BibTeX
@inproceedings{scharenborg2012hearing,
  author = {Odette Scharenborg and Esther Janse},
  title = {{Hearing loss and the use of acoustic cues in phonetic categorisation of fricatives}},
  booktitle = {Interspeech},
  year = {2012},
}
Figure