Related Experiment Video
Updated: Dec 30, 2025
09:30
Precision Induction and Distinction of Coughing and Sneezing Reflexes in Mice
Published on: October 3, 2025
715
SnoreNet: Detecting Snore Events from Raw Sound Recordings
Summary
A new deep learning model, SnoreNet, effectively detects snoring, an early sign of Obstructive Sleep Apnea Hypopnea Syndrome (OSAHS). This sound-based approach offers a non-invasive method for OSAHS analysis.
Area of Science:
- Biomedical Engineering
- Signal Processing
- Artificial Intelligence
Background:
- Snoring is a primary indicator of Obstructive Sleep Apnea Hypopnea Syndrome (OSAHS).
- Developing non-invasive, sound-based diagnostic tools for OSAHS is crucial.
- Automated snore detection is a key step towards accessible OSAHS analysis.
Purpose of the Study:
- To introduce SnoreNet, a novel deep neural network for accurate snore detection.
- To develop a sound-based method for identifying snoring in continuous audio recordings.
- To create an automated system that requires no manual feature engineering.
Main Methods:
- A deep neural network architecture, SnoreNet, was designed for snore detection.
- The model processes continuous sound recordings to identify snore events.
- SnoreNet utilizes multiple feature maps and default bounding boxes with varying scales to accommodate temporal variations in snoring.
Main Results:
- SnoreNet achieved an average precision (AP) of 81.82% on a newly collected dataset.
- The model demonstrated effectiveness in capturing snore characteristics without manual feature engineering.
- The deep learning approach successfully detected snores from audio recordings.
Conclusions:
- SnoreNet presents a simple yet effective deep learning solution for snore detection.
- The proposed method advances non-invasive, sound-based approaches for OSAHS analysis.
- This work contributes to the development of low-cost, automated tools for sleep apnea screening.
