Prediction of Sleep Stages Via Deep Learning Using Smartphone Audio Recordings in Home Environments: Model...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Sleep Science

Background:

• Public interest in sleep monitoring is increasing, driving demand for home-based solutions.
• Sound-based sleep staging using deep learning is a promising, convenient alternative to wearable devices.
• Previous studies used only in-laboratory data; home environments present challenges due to background noise.

Purpose of the Study:

• To develop and validate a deep learning method for sound-based sleep staging in uncontrolled home environments.
• To address the lack of annotated home sleep data by combining hospital and home recordings.
• To investigate the feasibility of using only smartphone audio for sleep staging at home.

Main Methods:

• A deep learning model was trained using hospital polysomnography (PSG) and audio data, augmented with home noise.
• Transfer learning incorporated smartphone audio recordings from home environments.
• Consistency training was applied using augmented hospital sound data to improve robustness.

Main Results:

• The model achieved 76.2% accuracy in sound-based sleep staging using home recordings.
• Performance was robust across various demographic groups and sleep apnea severities.
• Combining transfer learning and consistency training yielded the largest accuracy improvement (+7.0%).

Conclusions:

• Sound-based sleep staging is a viable option for home use.
• Advanced deep learning techniques enable accurate sleep stage prediction from smartphone audio in real-world settings.
• This approach eliminates the need for specialized equipment, making sleep monitoring more accessible.