Multifeature Fusion Method with Metaheuristic Optimization for Automated Voice Pathology Detection

Area of Science:

• Medical acoustics
• Computational linguistics
• Artificial Intelligence in Medicine

Background:

• Voice pathologies significantly impact speech and require accurate, timely diagnosis.
• Current diagnostic methods are often time-consuming, costly, and patient-unfriendly.
• Automated detection systems are needed for efficient screening and management of vocal disorders.

Purpose of the Study:

• To develop and evaluate a metaheuristic-based automated system for voice pathology detection.
• To enhance the accuracy and efficiency of vocal disorder diagnosis using machine learning.
• To reduce the reliance on manual acoustic analysis for voice pathology identification.

Main Methods:

• Extracted acoustic features (Zero-Crossing Rate, Root-Mean-Square Energy, Mel-frequency Cepstral Coefficients) from 1000 voice signals.
• Utilized a hybrid approach combining features and employed the Grey Wolf Optimizer (MELGWO) for feature map optimization.
• Applied supervised machine learning classifiers, including Support Vector Machine (SVM) and K-nearest neighbors, for pathology classification.

Main Results:

• The MELGWO algorithm effectively optimized feature maps, reducing computational load and improving model performance.
• The hybrid feature set and optimized parameters led to high diagnostic accuracy.
• The SVM classifier, using MELGWO-optimized features, achieved a peak accuracy of 99.50%.

Conclusions:

• Metaheuristic optimization significantly enhances automated voice pathology detection accuracy.
• The proposed automated system offers a promising, efficient, and accurate alternative to traditional diagnostic methods.
• This technology can facilitate earlier diagnosis, improve patient outcomes, and streamline vocal health monitoring.