Table 1.

The results table presents the publications included in this review study. The following information is summarized from left to right: first author, year of publication, journal, methods, outcomes of the study, validation method, and precision (as reported in the paper).

First Author	Year	Journal	Methods	Study Setting	Device	Outcomes	Validity	Precision
Cadavid [33]	2012	Pers. Ubiquit. Comput.	Active Appearance Model (AAM) for face tracking, and spectral analysis on the temporal window of the model parameter values; binary support vector machine classifier for chewing events	Laboratory	Not reported; 37 videos at 24 fps; frame resolution: 640 × 480	Chewing detection	Manual annotation for chewing events	93% after cross-validation
Okamoto [34]	2014	IEEE International Conference on Multimedia and Expo Workshops	Mouth detector limited to the lower part of detected face; Chopstick detection using OpenCV Hough transform for straight lines	Laboratory	Smartphone Google Nexus 5 (2.3 GHz Quad Core, Android 4.4), inner camera; frontal view	Food intake estimation	N.A.	N.A.
Hantke [35]	2018	Proceedings of the 20th ACM International Conference on Multimodal Interaction	OpenFace facial landmarks extraction for tracking the mouth	Office room	Logitech HD Pro Webcam C920; 30 fps; resolution: 1280 × 720; frontal view	Food liking	Leave-One-Out Cross-Validation and SVM	Likability 0.583
Haider [36]	2018	Proceedings of the 20th ACM International Conference on Multimodal Interaction	OpenSMILE for facial landmarks extraction, coupled with OpenSMILE audio-feature extraction	Office room	Logitech HD Pro Webcam C920; 30 fps; resolution: 1280 × 720; frontal view	Food liking	Leave-One-Out Cross-Validation and active feature transformation	0.61
Konstantinidis [37]	2019	Computer Vision Systems	OpenPose for mouth and hands tracking; Deep Network (3 Conv + shortcut, 3 Conv + shortcut, 3 LSTM)	Laboratory	85 videos; Samsung digital camcorder; 1.5 m away from the subject; side view	Automatic bite detection	F-Score: 0.9173	0.9175
Qiu [38]	2019	IEEE 16th International Conference on Wearable and Implantable Body Sensor Networks	Mask R-CNN for 360-degree camera meal videos; Thresholds for assessing pixel intersection between hand-face and hand-food to infer eating events	Free-living (Indoor food sharing scenarios)	Samsung’s gear 360 camera; 1024 × 1024 pixels	Food intake estimation	N.A.	N.A.
Hossain [39]	2020	IEEE Access	Face detection with manual selection of the region of interest; CNN for bite/non bite classification; Optical flow for spatial chewing motion at every pixel	Laboratory	84 videos; SJCAM SJ4000 Action Camera; 1080p video at 30 fps; side view	Automatic count of bites and chews	Manual annotation with 3-button system and LabView software (custom-made)	Bites: 88.9% ± 7.4%; Chews: 88.64% ± 5.29%
Rouast and Adam [40]	2020	IEEE J. Biomed. Health Inform	CNN for hand-to-mouth movement in 360-degree meal videos	Free-living (Indoor group meal)	102 videos; 360 fly-4 K camera; 24 fps	Intake gesture detection	N.A.	F1-score: 0.858
Konstantinidis [41]	2020	Nutrients	OpenPose skeletal and mouth features extracted for training the RABiD algorithm. Two stream data: 2D coordinates and distances from mouth corners, and from upper body	Laboratory	Samsung digital camcorder; 1.5 m away from the subject; side view; resolution: 576p (720 × 576 pixels) at 25 fps	Meal duration and bite counts	Manual annotation (Noldus Observer XT)	F1-score: 0.948
Nour [42]	2021	Advances in Social Sciences Research Journal	Facial landmarks (dlib) for tracking jawline movement; OpenPose for 2D pose estimation	N.A.	N.A.	Real-time eating activity tracking	Manual annotation	N.A.
Park [43]	2020	Robotics and Autonomous System	Facial landmarks (dlib) for mouth-pose estimator; Algorithmic model for improving 3D estimation, location, and orientation of the mouth	Laboratory	Intel SR300 RGB-D camera	Robot active feeding assistance	Wrist-mounted camera	N.A.
Alshboul [44]	2021	Sensors	Time series data consisting of Euclidean distance between jaw/mouth landmarks and a reference facial landmark	Free-living (outdoors, indoors, and public spaces)	300 videos; Huawei Y7 Prime 2018 smartphone; 13 MP camera; resolution: 1080p at 30 fps; frontal view	Number of chews	Manual annotation (Intra-class correlation coefficient = slow: 0.96, normal: 0.94, fast: 0.91)	Avg Error ± SD: 5.42% ± 4.61 (slow chewing) 7.47% ± 6.85 (normal chewing) 9.84% ± 9.55 (fast chewing)
Kato [45]	2021	Gerodontology	Video fluoroscopy of swallowing for determining which foods are more appropriate for elderly people	Laboratory	N.A.	Association between masticatory movements and food texture in older adults	N.A.	N.A.