A HTK-based Method for Detecting Vocal Fold Pathology

Area of Science:

• Speech analysis
• Medical diagnostics
• Acoustic phonetics

Background:

• Acoustic analysis methods for vocal fold pathology detection are emerging.
• Existing methods predominantly focus on phoneme-based detection.
• Detection based on continuous speech remains an under-researched area.

Purpose of the Study:

• To develop a novel method for vocal fold pathology detection using continuous speech.
• To apply this method to Russian digits for Belorussian populations.
• To evaluate the accuracy of the proposed detection technique.

Main Methods:

• Utilized the Hidden Markov Model Toolkit (HTK) for acoustic analysis.
• Developed a three-state Hidden Markov Model (HMM) for phoneme modeling.
• Applied the method to continuous Russian digits (1-10).

Main Results:

• The proposed method achieved a 90% detection accuracy rate.
• Experimental results demonstrate the effectiveness of the continuous speech approach.
• The method shows high potential for accurate pathology identification.

Conclusions:

• This work presents a pioneering approach to vocal fold pathology detection using continuous Russian digits.
• The high accuracy suggests its utility as an auxiliary diagnostic tool.
• The method is recommended for implementation in medical centers.