Table 2.

Summary of surface embedded literature. The drinking detection accuracy shows the classification accuracy for detecting drinking action only, while the system accuracy is the average classification accuracy considering all classes. The weight error/accuracy shows the performance for identifying the volume intake.

Ref.	#Sen	Method	#Sub	System Accuracy (%)	Drinking Detection Accuracy (%)	Weight Error/ Accuracy	Limitations
[87]	9+	Rule-based, template matching	3	80	-	82.62% accuracy	Small sample size, all objects need RFID
[89]	1264	DT 1, 7-class No LOSO 2	5	91	99	16% RMSE	Low weight accuracy
		With LOSO 2		76	99
[91]	1	Segmentation and thresholding	271	39% of bites are undetected	39% of drink sips undetected	-	Many false positives and undetected intakes
[92]	8	Comparing against acoustic neck microphone	2	-	-	<9 g error	Small sample size

¹ Decision Tree, ² Leave-one-subject-out