Snoring - An Acoustic Definition.

Area of Science:

• Acoustic analysis
• Signal processing
• Sleep medicine

Background:

• Distinguishing snoring from loud breathing is subjective.
• An objective definition based on acoustic parameters is needed.

Purpose of the Study:

• Identify and assess acoustic features for distinguishing snoring and loud breathing.
• Facilitate an objective definition of snoring using acoustic parameters.

Main Methods:

• Utilized a corpus of snore and breath sounds from 23 subjects.
• Employed the openSMILE feature extractor to evaluate 6,373 acoustic features.
• Assessed feature selectivity using Support Vector Machine (SVM) classification, logistic regression, and single-feature recall.

Main Results:

• Statistical functionals of mel frequency-generated perceptual linear predictive (PLP) cepstral coefficients achieved up to 93.8% unweighted average recall (UAR).
• Low-level descriptors (LLDs) and statistical functionals based on fast Fourier transformation (FFT) yielded a 93.0% UAR with SVM classification.
• Summed mel frequency spectrum-generated PLP cepstral coefficients achieved a 92.2% UAR using SVM classification.

Conclusions:

• Snoring and loud breathing can be objectively distinguished using robust acoustic features.
• Findings can guide the development of a consensus for an objective definition of snoring versus loud breathing.