A data augmentation-based technique to classify chewing and swallowing using LSTM
Muhammad Mehedi Billah, Masafumi Nishimura · 2020 IEEE 2nd Global Conference on Life Sciences and Technologies (LifeTech) · 2020
Obesity is currently a common problem for most countries. It can be controlled if a person's eating habits are monitored. Eating sounds such as, chewing and swallowing can be used to monitor individual eating habits. Numerous systems have already been developed to automatically classify chewing and swallowing by using information extracted from eating sounds. However, a large amount of well-labeled training data is required to achieve better accuracy, and manually labeling the eating sound data requires immense time. This study proposed a classification system based on the data augmentation technique to automatically generate a large amount of training data. The newly generated data could then be used along with the labeled data to train the long short-term memory (LSTM) network to classify chewing and swallowing. This experiment compared a system trained only with the labeled data and its results revealed the effectiveness of the proposed approach.