Pitch deviation analysis of pathological voice in connected speech

Area of Science:

• Speech and Hearing Sciences
• Biomedical Engineering
• Otolaryngology

Background:

• Distinguishing between normal and pathologic voices is crucial for diagnosing vocal fold pathologies.
• Traditional voice analysis methods often rely on sustained vowels, limiting their application in natural speech.
• A novel algorithm analyzing pitch deviation in connected speech offers a potential advancement.

Purpose of the Study:

• To compare the efficacy of a novel voice analysis algorithm in differentiating normal and pathologic voices during connected speech.
• To evaluate the clinical potential of this algorithm for voice disorder assessment.

Main Methods:

• Adult vocalizations from 10 normal subjects and 31 patients with benign vocal fold lesions were analyzed.
• A novel algorithm generated 2D patterns of pitch deviation from connected speech samples.
• Measures from the novel algorithm were compared to standard jitter and shimmer measures from sustained /a/ using the Computerized Speech Lab (CSL).

Main Results:

• Over 58% of measures from connected speech analysis outperformed standard CSL jitter and shimmer in discriminating between normal and abnormal voice populations.
• Twenty-five percent of experimental measures, including those from sustained /a/, showed statistically significant differences (p < .01%) between populations.

Conclusions:

• The novel voice analysis algorithm successfully distinguishes between normal and abnormal voice populations using connected speech samples.
• This algorithm demonstrates clinical potential as a tool for voice disorder diagnosis.