Acoustic analysis of intonation in parkinson's disease

Joan K. Y. Ma, Rüdiger HoffmannInterspeech 2010

TL;DR

This study investigates how Parkinson's Disease (PD) affects the way speakers mark different sentence types (statements, questions, imperatives) using intonation. While PD speakers show reduced pitch variation, they still manage to differentiate intonations similarly to healthy speakers, suggesting the core issue lies in motor control rather than cognitive understanding of intonation.

What Problem Does This Paper Address?

Parkinson's Disease (PD) is known to cause speech difficulties, particularly affecting prosody (the rhythm, stress, and intonation of speech). This condition, often leading to hypokinetic dysarthria, can make speech sound monotonous and less expressive. While it's generally accepted that PD impacts prosody, the specific ways in which PD speakers mark different sentence types (like distinguishing a statement from a question using intonation) at an acoustic level, and whether this is due to cognitive or motor deficits, remains an open question. Previous research has often focused on limited acoustic features, leading to inconclusive results. This paper aims to provide a more comprehensive acoustic analysis of intonation marking in PD speakers across multiple prosodic dimensions.

Methods

The researchers conducted an acoustic analysis of recorded speech samples from 24 German speakers with Parkinson's disease (mild to moderate severity) and 12 non-dysarthric control speakers. They analyzed three key prosodic dimensions:

  • Fundamental Frequency (F0): This relates to the pitch of the voice. They measured average F0, F0 range (the difference between the highest and lowest pitch), and F0 envelop (the variation in pitch within sentences and between syllables).
  • Intensity: This relates to the loudness of the voice. Similar to F0, they measured average intensity, intensity range, and intensity envelop.
  • Speech Rate: Measured as syllables per second.

To understand how PD affects intonation marking, speakers produced 15 German sentences, with five sentences each for statements, questions, and imperatives. The sentences varied in length. Statistical analysis, specifically a two-way ANOVA, was used to compare the acoustic measures between the speaker groups (mild PD, moderate PD, control) and across the different intonation types.

Datasets

The study used speech recordings from 24 native German speakers diagnosed with idiopathic Parkinson's disease and 12 native German control speakers. The PD speakers were recruited from the Clinic and Polyclinic of Neurology at Dresden University Hospital and were categorized as mild or moderate dysarthric based on a passage reading sample. Control speakers had no history of speech or language disorders. The speech material consisted of 15 German sentences (5 statements, 5 questions, 5 imperatives) of varying syllable counts (7, 9, 11, 13, 15) recorded in a quiet room. No new benchmark datasets were introduced; existing speech samples were analyzed.

Key Insights and Ideas

  • PD speakers are "monotonous" in pitch: PD speakers, regardless of severity, exhibited significantly higher average F0 and reduced F0 variability (narrower F0 range and smaller F0 envelop) compared to control speakers. This aligns with the known characteristic of "monopitch" in PD speech.
  • Motor deficit likely culprit for prosody issues: Despite reduced pitch variation, PD speakers were still able to differentiate between statements, questions, and imperatives acoustically, similar to control speakers. This suggests that the cognitive ability to understand and intend different intonations is preserved, but the motor execution of these nuances is impaired.
  • Intensity findings are complex: While mild PD speakers showed reduced intensity variability (smaller range and envelop), their average intensity was higher than controls. Moderate PD speakers' average intensity was comparable to controls. This finding is somewhat unexpected given the common report of "hypophonia" (reduced loudness) in PD, suggesting that speakers might compensate, especially in controlled speech tasks.
  • Speech rate is surprisingly consistent: No significant differences were found in how PD speakers and control speakers varied their speech rate across different intonation types. This implies that the timing aspect of intonation marking might be less affected than pitch.

Key Results

  • F0: PD speakers (mild and moderate) had significantly higher average F0 and significantly narrower F0 range and F0 envelop compared to control speakers.
  • Intensity: Mild PD speakers had significantly higher average intensity than moderate PD and control speakers. Mild PD speakers also showed significantly narrower intensity range and smaller intensity envelop than control speakers.
  • Speech Rate: No significant differences were found in the speech rate variation across intonations between the speaker groups.
  • Intonation Contrast: All speaker groups showed similar patterns of intonation contrast across the three sentence types for F0 and intensity measures, with questions generally having higher average F0 than statements.

Limitations and Future Work

Limitations: * The study focused on a specific set of acoustic parameters and did not explore all possible prosodic dimensions that might be relevant for intonation marking. * The intensity findings were somewhat unexpected and suggest that speakers might compensate for hypophonia in controlled speech tasks. The authors note that further research using different speech tasks might be beneficial. * The heterogeneity of PD speech characteristics is acknowledged, and the study's findings might not generalize to all PD speakers.

Future Work: * Investigate speech loudness in PD speakers using different speech tasks to better understand compensation strategies. * Explore a wider range of acoustic parameters and prosodic dimensions to capture a more complete prosodic profile of PD speech. * Conduct perceptual studies to correlate the acoustic findings with listener perception of intonation in PD speech.

Summary updated

BibTeX
@inproceedings{ma2010acoustic,
  author = {Joan K. Y. Ma and Rüdiger Hoffmann},
  title = {{Acoustic analysis of intonation in parkinson's disease}},
  booktitle = {Interspeech},
  year = {2010},
}
Figure