To realize ubiquitous eating habits monitoring, we proposed the use of chewing sound sensed by an in-ear placed wireless wearable microphone. A prototype of wireless wearable in-ear microphone was developed by utilizing a common Bluetooth headset. We proposed robust chewing number counting algorithm which consists of two recognition stages: "chew-like" signal detection and chewing verification stage. As a result, average chewing number counting error rate of 1.93% was achieved. Lastly, chewing sound mapping was proposed to provide an additional intuitive feedback for users to be able to infer the eating habits in their daily life context.
展开▼