TL;DR
This study investigates how individuals with spastic dysarthria, a motor speech disorder, produce lexical stress (the emphasis on a syllable within a word) compared to healthy speakers. The findings reveal that dysarthric speakers use acoustic cues differently, particularly for words with stress on the second syllable, suggesting a potential target for speech therapy to improve intelligibility.
What Problem Does This Paper Address?
Lexical stress is crucial in English for distinguishing word meanings (e.g., "CONtract" vs. "conTRACT") and is generally marked by variations in intensity, pitch (fo), and duration. However, individuals with dysarthria, often characterized by reduced motor control affecting speech, may struggle to produce these prosodic contrasts effectively. While previous research has explored sentential stress in dysarthria, the specific acoustic mechanisms of lexical stress production in this population, and how they are influenced by stress location within a word (first vs. second syllable), remain less understood. This paper aims to fill this gap by quantitatively analyzing the acoustic cues used by speakers with spastic dysarthria to mark lexical stress, comparing them to healthy controls.
Methods
The researchers recorded isolated word productions from both speakers with spastic dysarthria and healthy control speakers. They then analyzed these recordings using acoustic measures. Specifically, they focused on six acoustic features for the nucleus vowel of stressed and unstressed syllables: peak intensity, average intensity, integral intensity (RMS energy over duration), peak fo, average fo, and duration. To quantify the prominence of stressed syllables, they calculated ratios of these measures for the stressed syllable over the unstressed syllable. Statistical analyses, including two-way ANOVAs, were used to examine the effects of stress location (first vs. second syllable) and group (dysarthria vs. control) on these acoustic ratios, as well as their interaction.
Datasets
The speech data for speakers with dysarthria was a subset of the Universal Access (UA) database, which was developed for automatic speech recognition systems for dysarthric speech. This database contains recordings from individuals with cerebral palsy. Control speakers were recruited separately. The study used a total of 30 tri- or quadra-syllabic words per speaker, with 11 words having stress on the first syllable and 19 on the second.
Key Insights and Ideas
- Dysarthric speakers convey lexical stress differently: The core finding is that individuals with spastic dysarthria do not produce lexical stress in the same way as healthy speakers. This difference is particularly pronounced when the stress falls on the second syllable of a word.
- Differential reliance on acoustic cues: For words with stress on the first syllable, dysarthric speakers appear to rely more heavily on pitch and intensity cues compared to healthy speakers.
- Stress location matters significantly: The way acoustic cues are used to mark stress differs based on whether the stress is on the first or second syllable, and this pattern is not uniform between the dysarthric and control groups.
- Duration as a primary cue: For healthy speakers, duration appears to be a robust cue for lexical stress, especially for second-syllable stressed words. While dysarthric speakers also utilize duration, the findings suggest a potential compensatory strategy where they might lean more on pitch and intensity for first-syllable stress.
Key Results
- Peak Intensity: The dysarthria (DS) group showed a greater peak intensity ratio for first-syllable stressed words compared to second-syllable stressed words, while the control speakers (CS) showed the opposite pattern. Both groups differed significantly in peak intensity ratios between first and second-syllable stressed words, and between groups for both stress locations.
- Average Intensity: Similar to peak intensity, the DS group had a higher average intensity ratio for first-syllable stress, while the CS group showed the opposite. Significant differences were found between stress locations for both groups and between groups for second-syllable stress.
- Integral Intensity: A significant interaction was found. Only for the CS group did integral intensity ratios differ significantly between first and second-syllable stressed words. The DS and CS groups differed significantly for second-syllable stress.
- Pitch (fo): For first-syllable stressed words, the DS group showed a peak fo ratio close to 1, while the CS group's ratio was lower than 1, suggesting less modulation of pitch for first-syllable stress in controls. Both groups showed significant differences in peak fo ratios based on stress location and between groups. Average fo showed similar trends, with significant differences between groups for second-syllable stress.
- Duration: Duration ratios were greater than 1 for all conditions. For CS speakers, duration differences were more pronounced for second-syllable stress. The DS and CS groups differed significantly in duration ratios only for second-syllable stressed words.
Limitations and Future Work
The authors acknowledge that the study's findings are based on isolated word productions, which may not fully reflect the prosodic patterns in spontaneous speech. They suggest that a greater contribution of intensity and pitch might be observed in spontaneous speech, and this warrants future investigation. Additionally, they recommend future studies incorporating perceptual data to validate the acoustic findings and determine the perceptual relevance of different acoustic cues for lexical stress in dysarthric speech. The study also suggests that second-syllable stressed words might require more therapeutic focus to improve lexical stress marking in individuals with dysarthria.
Summary updated