Audio-based snore detection using deep neural networks.

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Artificial Intelligence

Background:

• Snoring is common and can indicate obstructive sleep apnea (OSA).
• Accurate snoring detection is crucial for OSA screening and diagnosis.

Purpose of the Study:

• To develop and evaluate a novel algorithm for accurate snore detection.
• To assess the impact of microphone placement on snore detection performance.

Main Methods:

• A hybrid CNN-RNN model was employed for snore detection.
• Audio recordings from 38 subjects with 5 strategically placed microphones were used.
• CNN extracted features from spectrograms; RNN classified snore events.

Main Results:

• The algorithm achieved high accuracy (95.3% ± 0.5%), sensitivity (92.2% ± 0.9%), and specificity (97.7% ± 0.4%).
• Optimal performance was observed with a microphone positioned 70 cm above the subject.
• Microphone placement variation showed minor differences in detection accuracy.

Conclusions:

• The CNN-RNN algorithm effectively detects snoring events with high accuracy.
• Microphone placement has a negligible effect on the overall performance of the snore detection system.