Hierarchical Classification and System Combination for Automatically Identifying Physiological and Neuromuscular...

Area of Science:

• Speech signal processing
• Laryngeal pathology diagnostics
• Computational linguistics

Background:

• Speech analysis aids in identifying voice disorders.
• Noninvasive methods are crucial for preliminary diagnosis of laryngeal pathologies.
• Accurate classification of voice pathologies is essential for effective treatment.

Purpose of the Study:

• To improve the accuracy of a three-class identification system for larynx pathologies.
• To evaluate the effectiveness of a hierarchical classifier combined with system analysis.
• To differentiate between healthy voices, physiological larynx pathologies, and neuromuscular larynx pathologies.

Main Methods:

• Utilized a hierarchical classification system combining sustained vowel /a/ and continuous speech analysis.
• Incorporated spectral and perceptual speech features, with and without formant information.
• Classified three groups: healthy (36), physiological pathologies (59), and neuromuscular pathologies (59).

Main Results:

• The combined system achieved an overall accuracy of 84.4%, a 9% improvement over standalone methods.
• Pathologic voice identification accuracy reached 98.7%.
• Identification accuracy for the two specific pathology classes was 81.3%.

Conclusions:

• Hierarchical classification and system combination offer significant advantages for classifying larynx pathologies.
• This modular approach enhances diagnostic capabilities for voice disorders.
• The findings support the use of advanced signal processing for improved laryngeal pathology identification.