Towards automated ingestion detection: swallow sounds.

Obesity is a worldwide epidemic and is a cause of many major chronic diseases. In most cases, obesity is a result of an imbalance between food intake and calories burned. Steps toward automated ingestion detection are being made. In order to automate the process of capturing ingestion, a method for detecting, analyzing, and recording sounds related to ingestion is being developed. In this paper, preliminary swallow sound analysis is presented and compared with various other noises captured from a throat mounted microphone. Initial frequency analysis indicates a stronger presence at high frequency intervals for swallow sounds in relation to other captured sounds such as voice. Comparisons show that a single high-pass filter can offer similar results as wavelet decomposition. Two simple methods for event detection are given.