Table 1.

Related complex activity recognition works.

Work	# Activities	Sensors	Details
Martínez-Pérez et al. [1]	4: taking blood pressure, feeding, hygiene, medication	RFID, accelerometers, video cameras	91.35% accuracy, 1 patient during 10 days. 81 instances.
Gu et al. [28]	26: making coffee, ironing, using phone, washing clothes, etc.	accelerometers, temperature, humidity, light, RFID, etc.	Overall accuracy 88.11%, 4 subjects over a 4 weeks period. Collected instances 532.
Cook et al. [29]	11: bathing, cooking, sleeping, eating, relaxing, taking medicine, hygiene, etc.	infrared motion detectors and magnetic door sensors	Accuracies of 71.08%, 59.76% and 84.89% for each of the 3 apartments during a period of 6 months.
Huynh et al. [30]	3: housework, morning tasks and shopping.	2D accelerometers and tilt switches	Accuracy of 91.8% for 1 user and period of about 10 h.
Kasteren et al. [35]	bathing, dressing, toileting, etc.	reed switches, pressure mats, mercury contacts, passive infrared, float sensors and temperature sensors	4 different datasets
Tolstikov et al. [38]	7: leaving, toileting, showering, sleeping, breakfast, etc.	14 binary sensors	Maximum accuracy of 95.7% for 1 subject during 27 days.
Vinh et al. [36]	4: dinner, commuting, lunch and office work	2 triaxial accelerometers	Precision of 88.47% for data collected during 7 days.
Sung et al. [39]	12: cooking, talking on the phone, working on computer, etc.	Microsoft Kinect	Average precision 86.5%, data collected by 4 subjects
Gordon et al. [40]	7: drinking, gesticulating, put mug on table, meeting, presentation, coffee break, etc.	accelerometers attached to mugs	Average accuracy of 95% for single-user and maximum 96% for group activities. 3 subjects. In total over 45 mins. of collected data.