Efficient snoring and breathing detection based on sub-band spectral statistics

Area of Science:

• Biomedical Engineering
• Signal Processing
• Sleep Medicine

Background:

• Snoring is a common symptom that may indicate obstructive sleep apnea (OSA).
• Accurate detection of snoring events is crucial for diagnosing sleep-related breathing disorders.
• Existing methods for snoring detection may require further refinement for improved accuracy.

Purpose of the Study:

• To propose and evaluate a novel method for detecting snoring events in sleep sound recordings.
• To analyze the acoustic characteristics of snoring sounds for feature extraction.
• To utilize machine learning for accurate classification of sleep-related audio events.

Main Methods:

• Developed a novel method analyzing acoustic properties of snoring sounds.
• Extracted feature vectors using mean and standard deviation of sub-band spectral energy.
• Employed a support vector machine (SVM) for frame-based classification of audio events.

Main Results:

• Achieved high accuracy in classifying sleep sound events: 99.61% for snoring, 99.16% for breathing, and 99.55% for silence.
• The method demonstrated effectiveness in distinguishing between snoring, breathing, and silence.
• Experimental validation used full-night audio recordings from individuals with snoring habits.

Conclusions:

• The proposed acoustic analysis and SVM-based method is highly effective for snoring detection.
• This technique offers a promising tool for objective assessment and diagnosis of sleep disorders like OSA.
• The high accuracy suggests potential for real-world clinical application in sleep monitoring.