Coherent Feature Extraction with Swarm Intelligence Based Hybrid Adaboost Weighted ELM Classification for Snoring...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Snoring is a common symptom of obstructive sleep apnea (OSA) and sleep-related breathing disorders, significantly impacting patient quality of life.
• Accurate snoring detection and classification are vital for OSA diagnosis, necessitating high-precision automated analysis algorithms.

Purpose of the Study:

• To develop and evaluate advanced algorithms for precise snoring sound analysis and classification.
• To compare the effectiveness of various feature extraction, selection, and classification techniques for snoring sounds.

Main Methods:

• Features were extracted from six domains: time, frequency, Discrete Wavelet Transform (DWT), sparse, eigenvalue, and cepstral.
• Feature selection was performed using Golden Eagle Optimization (GEO), Salp Swarm Algorithm (SSA), and Refined SSA.
• Classification employed eight traditional machine learning classifiers and two proposed hybrid models: Firefly Algorithm-Weighted Extreme Learning Machine with Adaboost (FA-WELM-Adaboost) and Capuchin Search Algorithm-Weighted Extreme Learning Machine with Adaboost (CSA-WELM-Adaboost).

Main Results:

• The best performance was achieved using DWT features, Refined SSA for feature selection, and the FA-WELM-Adaboost classifier, yielding an Unweighted Average Recall (UAR) of 74.23%.
• The second-best results were obtained with DWT features, GEO feature selection, and the CSA-WELM-Adaboost classifier, reporting a UAR of 73.86%.

Conclusions:

• Hybrid machine learning models, particularly FA-WELM-Adaboost and CSA-WELM-Adaboost, combined with DWT features and optimized selection techniques, show significant promise for accurate snoring sound classification.
• The developed methods offer a potential advancement in automated screening and diagnosis of obstructive sleep apnea through precise snoring analysis.