Chewing Behavior Detection Based on Facial Dynamic Features

Diet serves as the primary source of calorie intake for human beings, and maintaining a regular dietary intake is crucial for overall health. The pace or speed of chewing can significantly impact the body's response to food consumption. Traditionally, dietary monitoring has relied on manual assessment by clinicians, a process that is labor-intensive, time-consuming, and susceptible to inaccuracies. In this study, we introduce a novel image processing-based approach for quantitatively evaluating chewing and swallowing capabilities. In this method, facial recognition is employed to detect and calibrate facial features using the Dlib facial landmark model. This enables the precise identification of the mandible's position, facilitating the capture of the subject's chewing movements. Subsequently, signal processing techniques are applied to calculate the number of chewing instances. Experiments was conducted with five subjects of diverse genders and ages. The results indicated a mean absolute error of 6.48% in chewing count calculation. The proposed method offers the advantages of convenience and minimal error in comparison to similar studies.