Automatic Detection of Chewing and Swallowing Using Hybrid CTC/Attention
Akihiro Nakamura, K. Ohta, Takato Saito, Hiroshi Mineno, Daizo Ikeda, Masafumi Nishimura · 2020
Eating behavior is an important parameter representing health status. This paper evaluates the performance of a system with Hybrid CTC/Attention for monitoring eating behavior from the viewpoint of health maintenance. Previous studies have failed to detect the quality of eating behavior. Therefore, we examined the position of chewing (left/right/front) and the possibility of automatic detection of swallowing in order to grasp eating behavior in greater detail. We evaluated the chewing and swallowing of common foods and confirmed the model's stable detection performance.