Figure 1.

Hyoid bone tracking based on the sensor signals and dataset labelling. During a swallowing period, the neck vibration is sampled in the anterior–posterior (A–P), superior–interior (S–I) and medial–lateral (M–L) directions. In this study, the accelerometer shown in the figure is applied for all the patients. The deep learning architecture, SRNN is intended to track the hyoid bone with the informative features extracted from the multi-channel signals. The microphone shown in the figure is used for other purposes.