Towards automated ingestion detection: swallow sounds.

Area of Science:

• Biomedical Engineering
• Health Informatics
• Signal Processing

Background:

• Obesity is a global epidemic linked to chronic diseases, often caused by caloric imbalance.
• Automated monitoring of food intake is crucial for managing obesity.
• Current methods lack efficient, non-invasive ways to track ingestion events.

Purpose of the Study:

• To develop and evaluate a method for automated detection, analysis, and recording of ingestion sounds.
• To differentiate swallow sounds from other laryngeal noises.
• To explore signal processing techniques for accurate ingestion event identification.

Main Methods:

• A throat-mounted microphone was used to capture laryngeal sounds during ingestion.
• Frequency analysis was performed on recorded swallow sounds and other noises (e.g., voice).
• Comparison of signal processing techniques, including high-pass filtering and wavelet decomposition, for event detection.

Main Results:

• Swallow sounds exhibit a distinct high-frequency signature compared to voice and other laryngeal sounds.
• A simple high-pass filter demonstrated comparable efficacy to complex wavelet decomposition for swallow sound detection.
• Two straightforward event detection algorithms were proposed and evaluated.

Conclusions:

• Automated swallow sound analysis is a viable approach for ingestion monitoring.
• High-frequency characteristics of swallow sounds can be leveraged for simple and effective detection.
• This technology holds potential for obesity management and dietary behavior analysis.