Abstract
This study presents a snoring detection method based on hidden Markov models (HMMs) using a piezo snoring sensor. Snoring is a major symptom of obstructive sleep apnea (OSA). In most sleep studies, snoring is detected with a microphone. Since these studies analyze the acoustic properties of snoring, they need to acquire data at high sampling rates, so a large amount of data should be processed. Recently, several sleep studies have monitored snoring using a piezo snoring sensor. However, an automatic method for snoring detection using a piezo snoring sensor has not been reported in the literature. This study proposed the HMM-based method to detect snoring using this sensor, which is attached to the neck. The data from 21 patients with OSA were gathered for training and test sets. The short-time Fourier transform and short-time energy were computed so they could be applied to HMMs. The data were classified as snoring, noise and silence according to their HMMs. As a result, the sensitivity and the positive predictivity values were 93.3% and 99.1% for snoring detection, respectively. The results demonstrated that the method produced simple, portable and user-friendly detection tools that provide an alternative to the microphone-based method.
Original language | English (US) |
---|---|
Pages (from-to) | N41-N49 |
Journal | Physiological Measurement |
Volume | 34 |
Issue number | 5 |
DOIs | |
State | Published - May 2013 |
Externally published | Yes |
Keywords
- hidden Markov models
- piezo snoring sensor
- snoring detection
ASJC Scopus subject areas
- Biophysics
- Physiology
- Biomedical Engineering
- Physiology (medical)