Table 9.

Summary of research works that use ambient and wearable sensor technology.

Research Authors (Year)	Purpose	Characteristics	Outcomes	Sensors
Aghajan et al. [134] (2007)	Significant event detection	A sensor network that consisted of various types of sensors was used. Based on sensor data, event detection modalities with distributed processing were applied for smart home applications. More specifically, a distributed vision-based analysis was carried out for the detection of the occupant’s posture. Then, features from multiple cameras were combined via a rule-based approach for significant event detection.	96.7% accuracy.	Accelerometer sensors, video camera, PIR motion sensors.
Bang et al. [135] (2008)	Recognition of activities of daily living	An accelerometer and environmental-sensor-based approach was proposed. Conditional probabilities were used for recognition of daily activities that combine human motion and contacts with objects.	97% accuracy.	Accelerometer sensors, environmental sensors, PIR motion sensors.
Bianchi et al. [136] (2009)	Fall detection	The study was to evaluate barometric pressure along with accelerometer-based fall detection. Signal processing techniques (e.g., signal magnitude area) and a classification algorithm (support vector machines) were used to discriminate falls from typical daily activities.	97.5% accuracy.	Accelerometer sensors and barometric pressure sensors.
Cao et al. [137] (2009)	Recognition of activities of daily living	An event-driven context-aware computing model was proposed for recognizing daily activities.	Elderly health monitoring through the proposed system showed the effectiveness of the proposed model.	Video camera, accelerometer sensors.
Hein et al. [138] (2010)	Recognition of activities of daily living	A two-fold approach was described. First, sensors were selected based on interviews of elderly people, their relatives, and caregivers. Then, based on the outcome of the interviews, a sensor-based system was utilized to recognize different daily human activities.	Maximum 96.1% sensitivity and 90.3% specificity.	Accelerometer sensors, video camera, PIR motion sensors, door sensors.
Medjahed et al. [139] (2009)	Recognition of activities of daily living	A fuzzy-logic-based approach was proposed for robust human activity recognition on simulated data.	97% accuracy.	Sound sensors, PIR motion sensors, physiological sensors, state-change sensors.
Nyan et al. [140] (2006)	Fall detection	A fall detection approach was proposed using gyroscopes. Angles from different sides were explored for accurately modelling fall detection.	Maximum 100% sensitivity and 97.5% specificity.	Gyroscopes sensors, video camera.
Roy et al. [141] (2011)	Recognition of activities of daily living	This work proposes a framework of daily activity recognition that uses possibility theory and description logic-based semantic modeling. Different machine learning approaches (e.g., Gaussian mixture models, hidden Markov models, deep belief network) were analyzed.	95% accuracy.	Pressure sensors, accelerometer sensors, video sensors, PIR motion sensors.
Sim et al. [142] (2011)	Recognition of activities of daily living	The work applied mining of correlated patterns in activity recognition systems.	The correlated activity pattern mining approach showed 35.5% higher accuracy than typical frequent mining systems.	RFID sensors, accelerometer sensors, reed switches, PIR motion sensors, pressure sensors.
Srinivasan et al. [143] (2007)	Fall detection	The system applied triaxial accelerometer and motion detector sensor data in a two-step fall detection algorithm. First, the system tried to detect falls using the normalized energy expenditure from acceleration values. Then, falls were confirmed by considering the absence of motion. Some thresholds and logic were used to detect falls.	100% accuracy for coronal falls and 94.44% sagittal falls.	Accelerometer sensors, PIR motion sensors.
Tolkiehn et al. [144] (2011)	Fall detection	The system used a 3D accelerometer and a barometric pressure sensor for robust fall detection, along with detection of the fall direction. The basic probability-based amplitude and angular features were obtained from accelerometer sensors. Later, a pressure threshold was used.	Maximum 89.97% accuracy for fall prediction and 94.12% for fall direction.	Accelerometer sensor, barometric pressure sensor.